diff --git a/.claude/skills/add-jit-kernel/SKILL.md b/.claude/skills/add-jit-kernel/SKILL.md
index e63a7d77b..60c94f264 100644
--- a/.claude/skills/add-jit-kernel/SKILL.md
+++ b/.claude/skills/add-jit-kernel/SKILL.md
@@ -435,7 +435,7 @@ if torch.cuda.get_device_capability()[0] < 9:
 
 JIT kernel tests live under `python/sglang/jit_kernel/tests/`. **CI does not run `pytest` in that directory directly.** The unified runner `test/run_suite.py` discovers every `test_*.py` there (and every `bench_*.py` under `benchmark/`), collects `register_*_ci(...)` calls by **statically parsing each file's AST**, and executes the selected suite. Every test file must register at least one CUDA entry or the collector fails its sanity check.
 
-- **PR / per-commit CUDA suites** (see `test/run_suite.py` → `PER_COMMIT_SUITES`): JIT unit tests use `stage-b-kernel-unit-1-gpu-large` on H100 and `stage-b-kernel-unit-1-gpu-b200` on B200/SM100 paths (see `.github/workflows/pr-test-jit-kernel.yml`). Multi-GPU JIT tests use `stage-b-kernel-unit-8-gpu-h200`.
+- **PR / per-commit CUDA suites** (see `test/run_suite.py` → `PER_COMMIT_SUITES`): JIT unit tests use `base-b-kernel-unit-1-gpu-large` on H100 and `base-b-kernel-unit-1-gpu-b200` on B200/SM100 paths (see `.github/workflows/pr-test-jit-kernel.yml`). Multi-GPU JIT tests use `base-b-kernel-unit-8-gpu-h200`.
 - **Nightly kernel suite**: `nightly-kernel-1-gpu` with `--nightly` — typically used with `SGLANG_JIT_KERNEL_RUN_FULL_TESTS=1` in CI for expanded parameter grids (see `python/sglang/jit_kernel/utils.py` → `should_run_full_tests` / `get_ci_test_range`). Wired in `.github/workflows/nightly-test-nvidia.yml` (e.g. `python3 run_suite.py --hw cuda --suite nightly-kernel-1-gpu --nightly --continue-on-error`).
 
 Registration pattern (module level, **literal** `est_time` and `suite` strings — required for AST parsing):
@@ -443,9 +443,9 @@ Registration pattern (module level, **literal** `est_time` and `suite` strings 
 ```python
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=30, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=30, suite="base-b-kernel-unit-1-gpu-large")
 # Optional B200/SM100 registration for tests that cover Blackwell-specific code paths
-# register_cuda_ci(est_time=30, suite="stage-b-kernel-unit-1-gpu-b200")
+# register_cuda_ci(est_time=30, suite="base-b-kernel-unit-1-gpu-b200")
 # Optional second registration: same file also listed under the nightly kernel suite
 # register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 ```
@@ -457,9 +457,9 @@ Use `register_cuda_ci(..., disabled="reason")` if the file must stay in-tree but
 **Run like CI** (from repo root):
 
 ```bash
-(cd test && python3 run_suite.py --hw cuda --suite stage-b-kernel-unit-1-gpu-large)
+(cd test && python3 run_suite.py --hw cuda --suite base-b-kernel-unit-1-gpu-large)
 # For B200/SM100-specific coverage:
-(cd test && python3 run_suite.py --hw cuda --suite stage-b-kernel-unit-1-gpu-b200)
+(cd test && python3 run_suite.py --hw cuda --suite base-b-kernel-unit-1-gpu-b200)
 ```
 
 For fast iteration you can still run `pytest` on a single file locally; CI coverage is via `run_suite.py`.
@@ -472,7 +472,7 @@ import torch
 from sglang.jit_kernel.scale import scale
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=30, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=30, suite="base-b-kernel-unit-1-gpu-large")
 
 
 @pytest.mark.parametrize("dtype", [torch.float16, torch.bfloat16, torch.float32])
@@ -517,7 +517,7 @@ if __name__ == "__main__":
 
 ## Step 5: Add a benchmark (required)
 
-Benchmarks are `bench_*.py` files under `python/sglang/jit_kernel/benchmark/`. They are picked up by the same `run_suite.py` machinery as unit tests. Register them for **`stage-b-kernel-benchmark-1-gpu-large`** (PR JIT benchmark job: `python3 run_suite.py --hw cuda --suite stage-b-kernel-benchmark-1-gpu-large`).
+Benchmarks are `bench_*.py` files under `python/sglang/jit_kernel/benchmark/`. They are picked up by the same `run_suite.py` machinery as unit tests. Register them for **`base-b-kernel-benchmark-1-gpu-large`** (PR JIT benchmark job: `python3 run_suite.py --hw cuda --suite base-b-kernel-benchmark-1-gpu-large`).
 
 Create `python/sglang/jit_kernel/benchmark/bench_scale.py`:
 
@@ -537,7 +537,7 @@ from sglang.jit_kernel.benchmark.utils import (
 from sglang.jit_kernel.scale import scale as jit_scale
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=6, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=6, suite="base-b-kernel-benchmark-1-gpu-large")
 
 SIZE_LIST = get_benchmark_range(
     full_range=[2**n for n in range(10, 20)],  # 1K … 512K elements
@@ -585,7 +585,7 @@ python python/sglang/jit_kernel/benchmark/bench_scale.py
 Run the benchmark suite the way CI does:
 
 ```bash
-cd test && python3 run_suite.py --hw cuda --suite stage-b-kernel-benchmark-1-gpu-large
+cd test && python3 run_suite.py --hw cuda --suite base-b-kernel-benchmark-1-gpu-large
 ```
 
 ---
diff --git a/.claude/skills/ci-workflow-guide/SKILL.md b/.claude/skills/ci-workflow-guide/SKILL.md
index 99885d3ef..3b71881ca 100644
--- a/.claude/skills/ci-workflow-guide/SKILL.md
+++ b/.claude/skills/ci-workflow-guide/SKILL.md
@@ -11,7 +11,7 @@ This skill covers the CI **infrastructure** layer — how tests are dispatched,
 
 ## Naming Conventions
 
-- **Suite**: `stage-{a,b,c}-test-{gpu_count}-gpu-{hardware}` (e.g., `stage-b-test-1-gpu-small`)
+- **Suite**: `base-{a,b,c}-test-{gpu_count}-gpu-{hardware}` (e.g., `base-b-test-1-gpu-small`)
 - **Test group**: Directory-level registered test group under `test/registered/` (e.g., `hicache` maps to `test/registered/hicache/test_*.py`)
 - **CI runner**: `{gpu_count}-gpu-{hardware}` (e.g., `1-gpu-5090`, `4-gpu-h100`, `8-gpu-h200`)
 
@@ -23,7 +23,7 @@ This skill covers the CI **infrastructure** layer — how tests are dispatched,
 |------|------|
 | `.github/workflows/pr-test.yml` | Main workflow — all stages, jobs, conditions, matrix definitions |
 | `.github/workflows/pr-gate.yml` | PR gating: draft check, `run-ci` label, per-user rate limiting |
-| `.github/actions/check-stage-health/action.yml` | Cross-job fast-fail: queries API for any failed job |
+| `.github/actions/check-pr-test-health/action.yml` | Cross-job fast-fail: queries API for any failed job |
 | `.github/actions/wait-for-jobs/action.yml` | Stage gating: polls API until stage jobs complete |
 | `.github/actions/check-maintenance/action.yml` | Maintenance mode check |
 | `test/run_suite.py` | Suite runner: collects, filters, partitions, executes tests |
@@ -49,53 +49,53 @@ This skill covers the CI **infrastructure** layer — how tests are dispatched,
         │                                                     │
         ▼                                                     │
  ┌─────────────────────────────────────┐                      │
- │          Stage A (~3 min)           │                      │
+ │          Base A (~3 min)            │                      │
  │         pre-flight check            │                      │
  │                                     │                      │
  │  ┌─────────────────────────────┐    │                      │
- │  │ stage-a-test-1-gpu-small    │    │                      │
+ │  │ base-a-test-1-gpu-small    │    │                      │
  │  │ (small GPUs)                │    │                      │
  │  └─────────────────────────────┘    │                      │
  │  ┌─────────────────────────────┐    │                      │
- │  │ stage-a-test-cpu            │    │                      │
+ │  │ base-a-test-cpu            │    │                      │
  │  │ (CPU)                       │    │                      │
  │  └─────────────────────────────┘    │                      │
  └──────┬──────────────────────────────┘                      │
         │                                                     │
         ▼                                                     ▼
  ┌─────────────────────────────────────┐          ┌──────────────────────────┐
- │          Stage B (~30 min)          │          │      kernel test         │
- │           basic tests               │          └──────────────────────────┘
+ │          Base B (~30 min)           │          │      kernel test         │
+ │            base tests               │          └──────────────────────────┘
  │                                     │          ┌──────────────────────────┐
  │  ┌─────────────────────────────┐    │          │   multimodal gen test    │
- │  │ stage-b-test-1-gpu-small    │    │          └──────────────────────────┘
+ │  │ base-b-test-1-gpu-small    │    │          └──────────────────────────┘
  │  │ (small GPUs, e.g. 5090)     │    │
  │  └─────────────────────────────┘    │
  │  ┌─────────────────────────────┐    │
- │  │ stage-b-test-1-gpu-large    │    │
+ │  │ base-b-test-1-gpu-large    │    │
  │  │ (large GPUs, e.g. H100)     │    │
  │  └─────────────────────────────┘    │
  │  ┌─────────────────────────────┐    │
- │  │ stage-b-test-2-gpu-large    │    │
+ │  │ base-b-test-2-gpu-large    │    │
  │  │ (large GPUs, e.g. H100)     │    │
  │  └─────────────────────────────┘    │
  └──────┬──────────────────────────────┘
         │
         ▼
  ┌─────────────────────────────────────┐
- │          Stage C (~30 min)          │
+ │          Base C (~30 min)           │
  │          advanced tests             │
  │                                     │
  │  ┌─────────────────────────────┐    │
- │  │ stage-c-test-4-gpu-h100     │    │
+ │  │ base-c-test-4-gpu-h100     │    │
  │  │ (H100 GPUs)                 │    │
  │  └─────────────────────────────┘    │
  │  ┌─────────────────────────────┐    │
- │  │ stage-c-test-8-gpu-h200     │    │
+ │  │ base-c-test-8-gpu-h200     │    │
  │  │ (8 x H200 GPUs)             │    │
  │  └─────────────────────────────┘    │
  │  ┌─────────────────────────────┐    │
- │  │ stage-c-test-4-gpu-b200     │    │
+ │  │ base-c-test-4-gpu-b200     │    │
  │  │ (4 x B200 GPUs)             │    │
  │  └─────────────────────────────┘    │
  │  ┌─────────────────────────────┐    │
@@ -112,9 +112,9 @@ This skill covers the CI **infrastructure** layer — how tests are dispatched,
  └─────────────────────────────────────┘
 ```
 
-**Every stage test job** includes a `check-stage-health` step after checkout — if any job in the run has already failed, the job fast-fails (red X) with a root cause annotation.
+**Every stage test job** includes a `check-pr-test-health` step after checkout — if any job in the run has already failed, the job fast-fails (red X) with a root cause annotation.
 
-**Scheduled runs** skip `wait-for-stage-*` jobs, running all stages in parallel. Fast-fail is also disabled.
+**Scheduled runs** skip `wait-for-base-*` jobs, running all stages in parallel. Fast-fail is also disabled.
 
 ---
 
@@ -126,13 +126,13 @@ This skill covers the CI **infrastructure** layer — how tests are dispatched,
 |-------|-----------|-------------|----------------------|
 | **1. Test method → file** | `unittest -f` (failfast) | One test method fails → entire test file stops immediately | Yes |
 | **2. File → suite** | `run_unittest_files()` default | One test file fails → entire suite stops (`--continue-on-error` off) | Yes |
-| **3. Job → job (same stage)** | `check-stage-health` action | One job fails → other waiting jobs in same stage fast-fail (red X) | Yes |
-| **4. Stage → stage (cross-stage)** | `wait-for-stage` + `needs` | Stage A fails → stage B/C jobs skip entirely (never get a runner) | Yes (wait jobs skipped) |
+| **3. Job → job (same stage)** | `check-pr-test-health` action | One job fails → other waiting jobs in same stage fast-fail (red X) | Yes |
+| **4. Stage → stage (cross-stage)** | `wait-for-base-*` + `needs` | Base A fails → base B/C jobs skip entirely (never get a runner) | Yes (wait jobs skipped) |
 
 - **Layer 1**: `-f` flag appended to all `python3 -m pytest` / `unittest` invocations in `ci_utils.py`
 - **Layer 2**: `--continue-on-error` flag in `run_suite.py` — off for PRs, on for scheduled runs
-- **Layer 3**: `check-stage-health` auto-detects `schedule` event and skips; filters out cascade failures to show only root cause jobs
-- **Layer 4**: `wait-for-stage-*` jobs are conditioned on `github.event_name == 'pull_request'` — skipped for scheduled runs
+- **Layer 3**: `check-pr-test-health` auto-detects `schedule` event and skips; filters out cascade failures to show only root cause jobs
+- **Layer 4**: `wait-for-base-*` jobs are conditioned on `github.event_name == 'pull_request'` — skipped for scheduled runs
 
 ---
 
@@ -140,8 +140,8 @@ This skill covers the CI **infrastructure** layer — how tests are dispatched,
 
 | Aspect | PR (`pull_request`) | Scheduled (`cron`, every 6h) | `/rerun-stage` (`workflow_dispatch`) |
 |--------|---------------------|------------------------------|--------------------------------------|
-| **Stage ordering** | Sequential: A → B → C via `wait-for-stage-*` | Parallel (all at once) | Single target stage only |
-| **Cross-job fast-fail** | Yes (`check-stage-health`) | Yes | Yes |
+| **Stage ordering** | Sequential: A → B → C via `wait-for-base-*` | Parallel (all at once) | Single target stage only |
+| **Cross-job fast-fail** | Yes (`check-pr-test-health`) | Yes | Yes |
 | **continue-on-error** | No (stop at first failure within suite) | Yes (run all tests) | No |
 | **Retry** | Enabled | Enabled | Enabled |
 | **max_parallel** | 3 (default), 14 if `high priority` label | 14 | 3 (default), 14 if `high priority` |
@@ -152,23 +152,23 @@ This skill covers the CI **infrastructure** layer — how tests are dispatched,
 
 ## Stage Gating (`wait-for-jobs` action)
 
-`wait-for-stage-a` and `wait-for-stage-b` are lightweight `ubuntu-latest` jobs that poll the GitHub Actions API.
+`wait-for-base-a` and `wait-for-base-b` are lightweight `ubuntu-latest` jobs that poll the GitHub Actions API.
 
 **How it works:**
 1. Calls `listJobsForWorkflowRun` to list all jobs in the current run
-2. Matches jobs by exact name or prefix (for matrix jobs, e.g., `stage-b-test-1-gpu-small (3)`)
+2. Matches jobs by exact name or prefix (for matrix jobs, e.g., `base-b-test-1-gpu-small (3)`)
 3. If any matched job has `conclusion === 'failure'` → fail immediately (fast-fail)
 4. If all matched jobs are completed and count matches `expected_count` → success
 5. Otherwise → sleep `poll-interval-seconds` (default: 60s) and retry
-6. Timeout after `max-wait-minutes` (240 min for stage-a, 480 min for stage-b)
+6. Timeout after `max-wait-minutes` (240 min for base-a, 480 min for base-b)
 
-**Job specs example** (stage-b):
+**Job specs example** (base-b):
 ```json
 [
-  {"prefix": "stage-b-test-1-gpu-small", "expected_count": 8},
-  {"prefix": "stage-b-test-1-gpu-large", "expected_count": 14},
-  {"prefix": "stage-b-test-2-gpu-large", "expected_count": 4},
-  {"prefix": "stage-b-test-4-gpu-b200", "expected_count": 1}
+  {"prefix": "base-b-test-1-gpu-small", "expected_count": 8},
+  {"prefix": "base-b-test-1-gpu-large", "expected_count": 14},
+  {"prefix": "base-b-test-2-gpu-large", "expected_count": 4},
+  {"prefix": "base-b-test-4-gpu-b200", "expected_count": 1}
 ]
 ```
 
@@ -178,17 +178,17 @@ This skill covers the CI **infrastructure** layer — how tests are dispatched,
 
 ---
 
-## Cross-Job Fast-Fail (`check-stage-health` action)
+## Cross-Job Fast-Fail (`check-pr-test-health` action)
 
 Composite action called after checkout in every stage test job (21 jobs total across `pr-test.yml`, `pr-test-multimodal-gen.yml`, `pr-test-sgl-kernel.yml`, `pr-test-jit-kernel.yml`).
 
 **How it works:**
 1. Queries `listJobsForWorkflowRun` for the current workflow run
-2. Filters for **root cause failures only** — jobs with `conclusion === 'failure'` whose failing step is NOT `check-stage-health` (excludes cascade failures)
+2. Filters for **root cause failures only** — jobs with `conclusion === 'failure'` whose failing step is NOT `check-pr-test-health` (excludes cascade failures)
 3. If root cause failures found → calls `core.setFailed()` with the list of root cause job names
 4. If none → does nothing (step succeeds)
 
-**Cascade filtering**: When job A fast-fails due to health check, it also has `conclusion: failure`. Without filtering, job B would list both the original failure AND job A's fast-fail. The filter checks each failed job's `steps` array — if the failing step name contains `check-stage-health` or `Check stage health`, it's excluded from the root cause list.
+**Cascade filtering**: When job A fast-fails due to health check, it also has `conclusion: failure`. Without filtering, job B would list both the original failure AND job A's fast-fail. The filter checks each failed job's `steps` array — if the failing step name contains `check-pr-test-health` or `Check PR test health`, it's excluded from the root cause list.
 
 **Usage pattern:**
 ```yaml
@@ -197,8 +197,8 @@ steps:
     uses: actions/checkout@v4
     ...
 
-  - uses: ./.github/actions/check-stage-health
-    id: stage-health
+  - uses: ./.github/actions/check-pr-test-health
+    id: pr-test-health
 
   - name: Install dependencies        # skipped automatically if health check failed
     ...                                # (default if: success() is false)
@@ -213,7 +213,7 @@ steps:
 
 **Error message example:**
 ```
-Fast-fail: skipping — root cause job(s): stage-b-test-1-gpu-small (0), stage-b-test-1-gpu-small (1)
+Fast-fail: skipping — root cause job(s): base-b-test-1-gpu-small (0), base-b-test-1-gpu-small (1)
 ```
 
 ---
@@ -270,27 +270,26 @@ Large suites are split across matrix jobs using the **LPT (Longest Processing Ti
 
 | Suite | Partitions | Runner | max_parallel |
 |-------|-----------|--------|-------------|
-| `stage-a-test-1-gpu-small` | 1 (no matrix) | `1-gpu-5090` | — |
-| `stage-a-test-cpu` | 4 | `ubuntu-latest` | — |
-| `stage-b-test-1-gpu-small` | 8 | `1-gpu-5090` | 8 |
-| `stage-b-test-1-gpu-large` | 14 | `1-gpu-h100` | dynamic (3 or 14) |
-| `stage-b-test-2-gpu-large` | 4 | `2-gpu-h100` | — |
-| `stage-b-test-4-gpu-b200` | 1 (no matrix) | `4-gpu-b200` | — |
-| `stage-b-kernel-unit-1-gpu-large` | 1 (no matrix) | `1-gpu-h100` | — |
-| `stage-b-kernel-unit-1-gpu-b200` | 1 (no matrix) | `4-gpu-b200` | — |
-| `stage-b-kernel-unit-8-gpu-h200` | 1 (no matrix) | `8-gpu-h200` | — |
-| `stage-b-kernel-benchmark-1-gpu-large` | 1 (no matrix) | `1-gpu-h100` | — |
-| `stage-c-test-4-gpu-h100` | 3 | `4-gpu-h100` | — |
-| `stage-c-test-8-gpu-h200` | 4 | `8-gpu-h200` | — |
-| `stage-c-test-8-gpu-h20` | 2 | `8-gpu-h20` | — |
-| `stage-c-test-deepep-4-gpu-h100` | 1 (no matrix) | `4-gpu-h100` | — |
-| `stage-c-test-deepep-8-gpu-h200` | 1 (no matrix) | `8-gpu-h200` | — |
-| `stage-c-test-4-gpu-b200` | 3 | `4-gpu-b200` | — |
-| `stage-c-test-4-gpu-b200-small` | 3 | `4-gpu-b200-low-disk` | — |
-| `stage-c-test-8-gpu-b200` | registered only | `8-gpu-b200` | — |
-| `stage-c-test-4-gpu-gb200` | registered only | `4-gpu-gb200` | — |
+| `base-a-test-1-gpu-small` | 1 (no matrix) | `1-gpu-5090` | — |
+| `base-a-test-cpu` | 4 | `ubuntu-latest` | — |
+| `base-b-test-1-gpu-small` | 8 | `1-gpu-5090` | 8 |
+| `base-b-test-1-gpu-large` | 14 | `1-gpu-h100` | dynamic (3 or 14) |
+| `base-b-test-2-gpu-large` | 4 | `2-gpu-h100` | — |
+| `base-b-test-4-gpu-b200` | 1 (no matrix) | `4-gpu-b200` | — |
+| `base-b-kernel-unit-1-gpu-large` | 1 (no matrix) | `1-gpu-h100` | — |
+| `base-b-kernel-unit-1-gpu-b200` | 1 (no matrix) | `4-gpu-b200` | — |
+| `base-b-kernel-unit-8-gpu-h200` | 1 (no matrix) | `8-gpu-h200` | — |
+| `base-b-kernel-benchmark-1-gpu-large` | 1 (no matrix) | `1-gpu-h100` | — |
+| `base-c-test-4-gpu-h100` | 3 | `4-gpu-h100` | — |
+| `base-c-test-8-gpu-h200` | 4 | `8-gpu-h200` | — |
+| `base-c-test-8-gpu-h20` | 2 | `8-gpu-h20` | — |
+| `base-c-test-deepep-4-gpu-h100` | 1 (no matrix) | `4-gpu-h100` | — |
+| `base-c-test-4-gpu-b200` | 3 | `4-gpu-b200` | — |
+| `base-c-test-4-gpu-b200-small` | 3 | `4-gpu-b200-low-disk` | — |
+| `base-c-test-8-gpu-b200` | registered only | `8-gpu-b200` | — |
+| `base-c-test-4-gpu-gb200` | registered only | `4-gpu-gb200` | — |
 
-> **Note**: Kernel suites (`stage-b-kernel-*`) run via `pr-test-jit-kernel.yml` and `pr-test-sgl-kernel.yml`, not the main `pr-test.yml`. `stage-c-test-8-gpu-b200` is registered in `test/run_suite.py` but not wired to PR CI. The GB200 job is currently commented out in `pr-test.yml` until a company-owned runner is provisioned. Multimodal diffusion uses `python/sglang/multimodal_gen/test/run_suite.py`, not `test/run_suite.py`.
+> **Note**: Kernel suites (`base-b-kernel-*`) run via `pr-test-jit-kernel.yml` and `pr-test-sgl-kernel.yml`, not the main `pr-test.yml`. `base-c-test-8-gpu-b200` is registered in `test/run_suite.py` but not wired to PR CI. The GB200 job is currently commented out in `pr-test.yml` until a company-owned runner is provisioned. Multimodal diffusion uses `python/sglang/multimodal_gen/test/run_suite.py`, not `test/run_suite.py`.
 
 **Workflow usage:**
 ```yaml
@@ -298,7 +297,7 @@ strategy:
   matrix:
     partition: [0, 1, 2, 3, 4, 5, 6, 7]
 steps:
-  - run: python3 run_suite.py --hw cuda --suite stage-b-test-1-gpu-small \
+  - run: python3 run_suite.py --hw cuda --suite base-b-test-1-gpu-small \
            --auto-partition-id ${{ matrix.partition }} --auto-partition-size 8
 ```
 
@@ -320,7 +319,7 @@ Determines which test suites to run based on file changes.
 
 | Output | Triggers |
 |--------|----------|
-| `main_package` | Stage A/B/C test suites |
+| `main_package` | Base A/B/C test suites |
 | `sgl_kernel` | Kernel wheel builds + kernel test suites; also switches B200 jobs to kernel-build runner labels outside `target_stage` mode |
 | `jit_kernel` | JIT kernel test workflow |
 | `multimodal_gen` | Multimodal-gen test workflow |
@@ -348,10 +347,10 @@ group: pr-test-{event_name}-{branch}-{pr_sha}-{stage}
 
 ## How To: Add a New Stage Job
 
-1. Define the job in `pr-test.yml` with `needs: [check-changes, call-gate, wait-for-stage-X, ...]`
+1. Define the job in `pr-test.yml` with `needs: [check-changes, call-gate, wait-for-base-X, ...]`
 2. Copy the `if:` condition pattern from an existing same-stage job (handles `target_stage`, `schedule`, `main_package`)
 3. Add `checkout` step
-4. Add `check-stage-health` step (after checkout) — if any prior job failed, `core.setFailed()` fires and all subsequent steps auto-skip via default `if: success()`
+4. Add `check-pr-test-health` step (after checkout) — if any prior job failed, `core.setFailed()` fires and all subsequent steps auto-skip via default `if: success()`
 5. Add `check-maintenance` step
 6. Add `download-artifact` step if `sgl_kernel` changed
 7. Add `install dependencies` step
@@ -359,7 +358,7 @@ group: pr-test-{event_name}-{branch}-{pr_sha}-{stage}
 9. Add `upload-cuda-coredumps` step with `if: always()`
 10. Register the suite name in `PER_COMMIT_SUITES` in `test/run_suite.py`
 11. If using matrix, add `--auto-partition-id` and `--auto-partition-size` to the run command
-12. **Update `wait-for-stage-X`** job spec with the new job name and `expected_count` (if matrix)
+12. **Update `wait-for-base-X`** job spec with the new job name and `expected_count` (if matrix)
 13. **Add the job to `pr-test-finish.needs`** list
 
 ---
@@ -368,8 +367,8 @@ group: pr-test-{event_name}-{branch}-{pr_sha}-{stage}
 
 | Symptom | Likely cause | What to check |
 |---------|-------------|---------------|
-| All stage-B/C jobs green but steps skipped | Earlier job failed, `check-stage-health` triggered | Find the actual failed job (red X) |
-| `wait-for-stage-b` timeout | `expected_count` doesn't match matrix size | Verify job spec counts match `matrix:` array length |
+| All stage-B/C jobs green but steps skipped | Earlier job failed, `check-pr-test-health` triggered | Find the actual failed job (red X) |
+| `wait-for-base-b` timeout | `expected_count` doesn't match matrix size | Verify job spec counts match `matrix:` array length |
 | `pr-test-finish` fails but all jobs green | A job was `cancelled` (counts as failure in finish) | Check concurrency cancellation |
 | Tests pass locally but fail in CI | Partition assignment, runner GPU type, or `est_time` inaccuracy | Check which partition the test lands in; verify runner label |
 | Flaky test retried and passed | Retriable failure (accuracy/perf) | Check `[CI Retry]` markers in job logs |
diff --git a/.claude/skills/write-sglang-test/SKILL.md b/.claude/skills/write-sglang-test/SKILL.md
index 8bd49a8b6..3ed22137d 100644
--- a/.claude/skills/write-sglang-test/SKILL.md
+++ b/.claude/skills/write-sglang-test/SKILL.md
@@ -27,10 +27,10 @@ JIT kernel exception:
 
 | Scenario | Model | CI Registration | Suite |
 |----------|-------|-----------------|-------|
-| **Unit tests** (no server / engine launch) | None | `register_cpu_ci` (prefer) or `register_cuda_ci` | `stage-a-test-cpu` or `stage-b-test-1-gpu-small` |
-| **Common / backend-independent** (middleware, abort, routing, config, arg parsing) | `DEFAULT_SMALL_MODEL_NAME_FOR_TEST` (1B) | `register_cuda_ci` only | `stage-b-test-1-gpu-small` |
-| **Model-agnostic functionality** (sampling, session, OpenAI API features) | `DEFAULT_SMALL_MODEL_NAME_FOR_TEST` (1B) | `register_cuda_ci` (+ AMD if relevant) | `stage-b-test-1-gpu-small` |
-| **General performance** (single node, no spec/DP/parallelism) | `DEFAULT_MODEL_NAME_FOR_TEST` (8B) | `register_cuda_ci` | `stage-b-test-1-gpu-large` |
+| **Unit tests** (no server / engine launch) | None | `register_cpu_ci` (prefer) or `register_cuda_ci` | `base-a-test-cpu` or `base-b-test-1-gpu-small` |
+| **Common / backend-independent** (middleware, abort, routing, config, arg parsing) | `DEFAULT_SMALL_MODEL_NAME_FOR_TEST` (1B) | `register_cuda_ci` only | `base-b-test-1-gpu-small` |
+| **Model-agnostic functionality** (sampling, session, OpenAI API features) | `DEFAULT_SMALL_MODEL_NAME_FOR_TEST` (1B) | `register_cuda_ci` (+ AMD if relevant) | `base-b-test-1-gpu-small` |
+| **General performance** (single node, no spec/DP/parallelism) | `DEFAULT_MODEL_NAME_FOR_TEST` (8B) | `register_cuda_ci` | `base-b-test-1-gpu-large` |
 | **Bigger features** (spec, DP, TP, disaggregation) | Case by case | Case by case | See suite table below |
 
 **Key principle for E2E tests**: Do NOT add `register_amd_ci` unless the test specifically exercises AMD/ROCm code paths. Common E2E tests just need any GPU to run — duplicating across backends wastes CI time with no extra coverage.
@@ -50,7 +50,7 @@ Defined in `python/sglang/test/test_utils.py`:
 
 ### Naming Conventions
 
-- **Suite**: `stage-{a,b,c}-test-{gpu_count}-gpu-{hardware}` (e.g., `stage-b-test-1-gpu-small`)
+- **Suite**: `base-{a,b,c}-test-{gpu_count}-gpu-{hardware}` (e.g., `base-b-test-1-gpu-small`)
 - **CI runner**: `{gpu_count}-gpu-{hardware}` (e.g., `1-gpu-5090`, `4-gpu-h100`, `8-gpu-h200`)
 
 ### All CI Suites
@@ -59,25 +59,24 @@ Defined in `python/sglang/test/test_utils.py`:
 
 | Suite | Runner (label) | Description |
 |-------|----------------|-------------|
-| `stage-a-test-1-gpu-small` | `1-gpu-5090` | Quick checks on a small NVIDIA GPU before heavier stages |
-| `stage-a-test-cpu` | `ubuntu-latest` | CPU-only unit tests |
-| `stage-b-test-1-gpu-small` | `1-gpu-5090` | Core engine tests that fit a 5090-class card |
-| `stage-b-test-1-gpu-large` | `1-gpu-h100` | Tests that need H100-class memory or kernels (e.g. FA3) |
-| `stage-b-test-2-gpu-large` | `2-gpu-h100` | Two-GPU correctness and parallelism (TP/PP) on H100 |
-| `stage-b-test-4-gpu-b200` | `4-gpu-b200` | Early Blackwell coverage (SM100+ paths) on four GPUs |
-| `stage-b-kernel-unit-1-gpu-large` | `1-gpu-h100` | JIT kernel correctness tests under `python/sglang/jit_kernel/tests/` |
-| `stage-b-kernel-unit-1-gpu-b200` | `4-gpu-b200` | JIT kernel correctness tests for Blackwell / SM100-specific paths |
-| `stage-b-kernel-unit-8-gpu-h200` | `8-gpu-h200` | Multi-GPU JIT kernel correctness tests under `python/sglang/jit_kernel/tests/` |
-| `stage-b-kernel-benchmark-1-gpu-large` | `1-gpu-h100` | JIT kernel benchmark files under `python/sglang/jit_kernel/benchmark/` |
-| `stage-c-test-4-gpu-h100` | `4-gpu-h100` | Large 4-GPU H100 integration and scaling tests |
-| `stage-c-test-8-gpu-h200` | `8-gpu-h200` | Large 8-GPU H200 runs for big models and parallelism |
-| `stage-c-test-8-gpu-h20` | `8-gpu-h20` | Large 8-GPU H20 runs for big models |
-| `stage-c-test-deepep-4-gpu-h100` | `4-gpu-h100` | DeepEP expert-parallel and networking on four H100s |
-| `stage-c-test-deepep-8-gpu-h200` | `8-gpu-h200` | DeepEP at 8-GPU H200 scale |
-| `stage-c-test-8-gpu-b200` | `8-gpu-b200` | 8-GPU B200 suite (registered but not yet wired to a workflow) |
-| `stage-c-test-4-gpu-b200` | `4-gpu-b200` | 4-GPU B200 suite for large models on Blackwell |
-| `stage-c-test-4-gpu-b200-small` | `4-gpu-b200` | Smaller 4-GPU B200 suite split onto low-disk B200 runners |
-| `stage-c-test-4-gpu-gb200` | `4-gpu-gb200` | 4-GPU GB200 suite for Grace Blackwell; registered in `run_suite.py`, but the PR workflow is currently disabled until a runner is provisioned |
+| `base-a-test-1-gpu-small` | `1-gpu-5090` | Quick checks on a small NVIDIA GPU before heavier stages |
+| `base-a-test-cpu` | `ubuntu-latest` | CPU-only unit tests |
+| `base-b-test-1-gpu-small` | `1-gpu-5090` | Core engine tests that fit a 5090-class card |
+| `base-b-test-1-gpu-large` | `1-gpu-h100` | Tests that need H100-class memory or kernels (e.g. FA3) |
+| `base-b-test-2-gpu-large` | `2-gpu-h100` | Two-GPU correctness and parallelism (TP/PP) on H100 |
+| `base-b-test-4-gpu-b200` | `4-gpu-b200` | Early Blackwell coverage (SM100+ paths) on four GPUs |
+| `base-b-kernel-unit-1-gpu-large` | `1-gpu-h100` | JIT kernel correctness tests under `python/sglang/jit_kernel/tests/` |
+| `base-b-kernel-unit-1-gpu-b200` | `4-gpu-b200` | JIT kernel correctness tests for Blackwell / SM100-specific paths |
+| `base-b-kernel-unit-8-gpu-h200` | `8-gpu-h200` | Multi-GPU JIT kernel correctness tests under `python/sglang/jit_kernel/tests/` |
+| `base-b-kernel-benchmark-1-gpu-large` | `1-gpu-h100` | JIT kernel benchmark files under `python/sglang/jit_kernel/benchmark/` |
+| `base-c-test-4-gpu-h100` | `4-gpu-h100` | Large 4-GPU H100 integration and scaling tests |
+| `base-c-test-8-gpu-h200` | `8-gpu-h200` | Large 8-GPU H200 runs for big models and parallelism |
+| `base-c-test-8-gpu-h20` | `8-gpu-h20` | Large 8-GPU H20 runs for big models |
+| `base-c-test-deepep-4-gpu-h100` | `4-gpu-h100` | DeepEP expert-parallel and networking on four H100s |
+| `base-c-test-8-gpu-b200` | `8-gpu-b200` | 8-GPU B200 suite (registered but not yet wired to a workflow) |
+| `base-c-test-4-gpu-b200` | `4-gpu-b200` | 4-GPU B200 suite for large models on Blackwell |
+| `base-c-test-4-gpu-b200-small` | `4-gpu-b200` | Smaller 4-GPU B200 suite split onto low-disk B200 runners |
+| `base-c-test-4-gpu-gb200` | `4-gpu-gb200` | 4-GPU GB200 suite for Grace Blackwell; registered in `run_suite.py`, but the PR workflow is currently disabled until a runner is provisioned |
 
 #### Per-commit (AMD)
 
@@ -130,12 +129,12 @@ Nightly suites are listed in `NIGHTLY_SUITES` in [`test/run_suite.py`](../../../
 
 Use the lightest suite that meets your test's needs:
 
-- **No GPU required** → `stage-a-test-cpu`
-- **Most small GPU tests** → `stage-b-test-1-gpu-small` (default choice)
-- **Need H100 memory or Hopper features** → `stage-b-test-1-gpu-large`
-- **JIT kernel correctness** → `stage-b-kernel-unit-1-gpu-large`
-- **JIT kernel correctness for B200 / SM100 paths** → `stage-b-kernel-unit-1-gpu-b200`
-- **JIT kernel benchmarks** → `stage-b-kernel-benchmark-1-gpu-large`
+- **No GPU required** → `base-a-test-cpu`
+- **Most small GPU tests** → `base-b-test-1-gpu-small` (default choice)
+- **Need H100 memory or Hopper features** → `base-b-test-1-gpu-large`
+- **JIT kernel correctness** → `base-b-kernel-unit-1-gpu-large`
+- **JIT kernel correctness for B200 / SM100 paths** → `base-b-kernel-unit-1-gpu-b200`
+- **JIT kernel benchmarks** → `base-b-kernel-benchmark-1-gpu-large`
 - **Multi-GPU** → only when the test actually needs multiple GPUs
 
 ---
@@ -156,7 +155,7 @@ from sglang.srt.<module> import TargetClass
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=5, suite="base-a-test-cpu")
 # Prefer CPU. Only use register_cuda_ci when the test truly needs a GPU.
 
 class TestTargetClass(CustomTestCase):
@@ -196,7 +195,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=60, suite="stage-b-test-1-gpu-small")
+register_cuda_ci(est_time=60, suite="base-b-test-1-gpu-small")
 
 
 class TestMyFeature(CustomTestCase):
@@ -246,7 +245,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=300, suite="stage-b-test-1-gpu-large")
+register_cuda_ci(est_time=300, suite="base-b-test-1-gpu-large")
 
 
 class TestMyFeaturePerf(CustomTestCase):
@@ -321,21 +320,21 @@ from sglang.test.ci.ci_register import (
 )
 
 # Per-commit test (small 1-gpu, runs on 5090)
-register_cuda_ci(est_time=80, suite="stage-b-test-1-gpu-small")
+register_cuda_ci(est_time=80, suite="base-b-test-1-gpu-small")
 
 # Per-commit test (large 1-gpu, runs on H100)
-register_cuda_ci(est_time=120, suite="stage-b-test-1-gpu-large")
+register_cuda_ci(est_time=120, suite="base-b-test-1-gpu-large")
 
 # Nightly-only test
 register_cuda_ci(est_time=200, suite="nightly-1-gpu", nightly=True)
 
 # Multi-backend test (only when testing backend-specific code paths)
-register_cuda_ci(est_time=80, suite="stage-a-test-1-gpu-small")
+register_cuda_ci(est_time=80, suite="base-a-test-1-gpu-small")
 register_amd_ci(est_time=120, suite="stage-a-test-1-gpu-small-amd")
 register_npu_ci(est_time=400, suite="nightly-8-npu-a3", nightly=True)
 
 # Temporarily disabled test
-register_cuda_ci(est_time=80, suite="stage-b-test-1-gpu-small", disabled="flaky - see #12345")
+register_cuda_ci(est_time=80, suite="base-b-test-1-gpu-small", disabled="flaky - see #12345")
 ```
 
 Parameters:
@@ -354,12 +353,12 @@ JIT kernel files live outside `test/registered/` but still use registration:
 from sglang.test.ci.ci_register import register_cuda_ci
 
 # Correctness tests in python/sglang/jit_kernel/tests/
-register_cuda_ci(est_time=30, suite="stage-b-kernel-unit-1-gpu-large")
-register_cuda_ci(est_time=30, suite="stage-b-kernel-unit-1-gpu-b200")
-register_cuda_ci(est_time=120, suite="stage-b-kernel-unit-8-gpu-h200")
+register_cuda_ci(est_time=30, suite="base-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=30, suite="base-b-kernel-unit-1-gpu-b200")
+register_cuda_ci(est_time=120, suite="base-b-kernel-unit-8-gpu-h200")
 
 # Benchmarks in python/sglang/jit_kernel/benchmark/
-register_cuda_ci(est_time=6, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=6, suite="base-b-kernel-benchmark-1-gpu-large")
 
 # Optional nightly registration
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
diff --git a/.github/actions/check-stage-health/action.yml b/.github/actions/check-pr-test-health/action.yml
similarity index 89%
rename from .github/actions/check-stage-health/action.yml
rename to .github/actions/check-pr-test-health/action.yml
index af5dc376d..d919373ef 100644
--- a/.github/actions/check-stage-health/action.yml
+++ b/.github/actions/check-pr-test-health/action.yml
@@ -1,4 +1,4 @@
-name: Check Stage Health
+name: Check PR Test Health
 description: Fail fast if any job in the current workflow run has already failed, or if the lint check (from lint.yml) has failed. Auto-skips for scheduled runs. The jobs-failed check (but not the lint check) is bypassed when the PR carries the `bypass-fastfail` label.
 
 inputs:
@@ -10,16 +10,16 @@ inputs:
 runs:
   using: composite
   steps:
-    - name: Check stage health
+    - name: Check PR test health
       uses: actions/github-script@v7
       env:
-        SKIP_STAGE_HEALTH_CHECK: ${{ env.SKIP_STAGE_HEALTH_CHECK }}
+        SKIP_PR_TEST_HEALTH_CHECK: ${{ env.SKIP_PR_TEST_HEALTH_CHECK }}
       with:
         github-token: ${{ inputs.github-token }}
         script: |
           // Skip when explicitly requested via env var (e.g. release branch cut)
-          if (process.env.SKIP_STAGE_HEALTH_CHECK === 'true') {
-            core.info('Skipping health check (SKIP_STAGE_HEALTH_CHECK=true)');
+          if (process.env.SKIP_PR_TEST_HEALTH_CHECK === 'true') {
+            core.info('Skipping health check (SKIP_PR_TEST_HEALTH_CHECK=true)');
             return;
           }
 
@@ -83,12 +83,12 @@ runs:
             // match so we cover both inline + reusable forms without confusing
             // 'h20' with the 'h200' prefix.
             const baseName = j.name.split(/[ /]/)[0];
-            if (baseName === 'stage-c-test-8-gpu-h20') {
+            if (baseName === 'base-c-test-8-gpu-h20') {
               return false;
             }
             // If the failing step is the health check, it's a cascade — skip it
             const failedStep = (j.steps || []).find(s => s.conclusion === 'failure');
-            if (failedStep && (failedStep.name.includes('check-stage-health') || failedStep.name.includes('Check stage health'))) {
+            if (failedStep && (failedStep.name.includes('check-pr-test-health') || failedStep.name.includes('Check PR test health'))) {
               return false;
             }
             return true;
diff --git a/.github/actions/wait-for-jobs/action.yml b/.github/actions/wait-for-jobs/action.yml
index 240deb0c4..609853713 100644
--- a/.github/actions/wait-for-jobs/action.yml
+++ b/.github/actions/wait-for-jobs/action.yml
@@ -3,12 +3,12 @@ description: Poll and wait for specified jobs in the current workflow run to com
 
 inputs:
   stage-name:
-    description: 'Human-readable stage name for log messages (e.g. "stage-a")'
+    description: 'Human-readable stage name for log messages (e.g. "base-a")'
     required: true
   jobs:
     description: |
       JSON array of job specs to wait for. Each element is either:
-        - a string: exact job name (e.g. "stage-a-test-1-gpu-small")
+        - a string: exact job name (e.g. "base-a-test-1-gpu-small")
         - an object { "prefix": "...", "expected_count": N }: for matrix jobs
     required: true
   max-wait-minutes:
diff --git a/.github/workflows/_pr-test-sgl-kernel-build.yml b/.github/workflows/_pr-test-sgl-kernel-build.yml
index e7b57d7ed..75981dc16 100644
--- a/.github/workflows/_pr-test-sgl-kernel-build.yml
+++ b/.github/workflows/_pr-test-sgl-kernel-build.yml
@@ -24,7 +24,7 @@ on:
       git_ref:
         type: string
         default: ''
-      skip_stage_health_check:
+      skip_pr_test_health_check:
         description: 'Forwarded from the caller workflow input of the same name. Not consumed inside this reusable workflow today, but mirrored so the resulting env context matches the inline-job baseline.'
         type: boolean
         default: false
@@ -39,7 +39,7 @@ env:
   SGLANG_IS_IN_CI: true
   SGLANG_CUDA_COREDUMP: "1"
   SGLANG_JIT_DEEPGEMM_FAST_WARMUP: true
-  SKIP_STAGE_HEALTH_CHECK: ${{ inputs.skip_stage_health_check && 'true' || 'false' }}
+  SKIP_PR_TEST_HEALTH_CHECK: ${{ inputs.skip_pr_test_health_check && 'true' || 'false' }}
   FORCE_REBUILD_DEEPEP: '1'
   PR_TEST_BYPASS_MAINTENANCE_ON_MAIN: ${{ github.ref == 'refs/heads/main' && 'true' || 'false' }}
   USE_VENV: false
diff --git a/.github/workflows/_pr-test-stage.yml b/.github/workflows/_pr-test-stage.yml
index 90f606658..e544224d6 100644
--- a/.github/workflows/_pr-test-stage.yml
+++ b/.github/workflows/_pr-test-stage.yml
@@ -4,7 +4,7 @@ name: PR Test Stage
 # `caller_inputs` = toJson(inputs) as bundles. `partitions` is forwarded
 # separately to keep matrix expressions single-fromJson — its value is itself
 # a JSON string, so reading it via `check_changes` would need double fromJson,
-# and matrix can't use step-resolved values. stage-a-test-cpu stays inline in
+# and matrix can't use step-resolved values. base-a-test-cpu stays inline in
 # pr-test.yml (bespoke uv pip / protoc / rust-cache install).
 
 on:
@@ -23,7 +23,7 @@ on:
         type: string
         required: true
       caller_inputs:
-        description: 'toJson(inputs) from pr-test.yml. Read via fromJson(...).git_ref / skip_stage_health_check / test_parallel_dispatch.'
+        description: 'toJson(inputs) from pr-test.yml. Read via fromJson(...).git_ref / skip_pr_test_health_check / test_parallel_dispatch.'
         type: string
         required: true
       partitions:
@@ -62,7 +62,7 @@ env:
   SGLANG_IS_IN_CI: true
   SGLANG_CUDA_COREDUMP: "1"
   SGLANG_JIT_DEEPGEMM_FAST_WARMUP: true
-  SKIP_STAGE_HEALTH_CHECK: ${{ fromJson(inputs.caller_inputs).skip_stage_health_check && 'true' || 'false' }}
+  SKIP_PR_TEST_HEALTH_CHECK: ${{ fromJson(inputs.caller_inputs).skip_pr_test_health_check && 'true' || 'false' }}
   FORCE_REBUILD_DEEPEP: '1'
   PR_TEST_BYPASS_MAINTENANCE_ON_MAIN: ${{ github.ref == 'refs/heads/main' && 'true' || 'false' }}
   USE_VENV: false
@@ -97,7 +97,7 @@ jobs:
       - name: Export rdma_devices to job env
         run: echo "SGLANG_CI_RDMA_ALL_DEVICES=${{ steps.rc.outputs.rdma_devices || '' }}" >> "$GITHUB_ENV"
 
-      - uses: ./.github/actions/check-stage-health
+      - uses: ./.github/actions/check-pr-test-health
 
       - uses: ./.github/actions/check-maintenance
 
diff --git a/.github/workflows/ci-auto-bisect.yml b/.github/workflows/ci-auto-bisect.yml
index 3c77b5f1e..6de26745f 100644
--- a/.github/workflows/ci-auto-bisect.yml
+++ b/.github/workflows/ci-auto-bisect.yml
@@ -2,7 +2,7 @@ name: CI Auto Bisect
 
 on:
   workflow_run:
-    workflows: ["PR Test"]
+    workflows: ["PR Test Base"]
     types: [completed]
     branches: [main]
   workflow_dispatch: {}
diff --git a/.github/workflows/pr-states.yml b/.github/workflows/pr-states.yml
index 2227237f3..dffc21453 100644
--- a/.github/workflows/pr-states.yml
+++ b/.github/workflows/pr-states.yml
@@ -85,7 +85,7 @@ jobs:
               '---',
               '### CI States',
               '',
-              `Latest PR Test: ${ptStart}${ptText}${ptEnd}`,
+              `Latest PR Test (Base): ${ptStart}${ptText}${ptEnd}`,
               `Latest PR Test (Extra): ${peStart}${peText}${peEnd}`,
               outerEnd,
             ].join('\n');
diff --git a/.github/workflows/pr-test-extra.yml b/.github/workflows/pr-test-extra.yml
index 12212f894..8b8aa4bd7 100644
--- a/.github/workflows/pr-test-extra.yml
+++ b/.github/workflows/pr-test-extra.yml
@@ -36,8 +36,8 @@ on:
         required: false
         type: boolean
         default: false
-      skip_stage_health_check:
-        description: "Skip stage health check fast-fail (e.g. for release branch cuts)"
+      skip_pr_test_health_check:
+        description: "Skip PR test health check fast-fail (e.g. for release branch cuts)"
         required: false
         type: boolean
         default: false
@@ -50,7 +50,7 @@ env:
   SGLANG_IS_IN_CI: true
   SGLANG_CUDA_COREDUMP: "1"
   SGLANG_JIT_DEEPGEMM_FAST_WARMUP: true
-  SKIP_STAGE_HEALTH_CHECK: ${{ (inputs.skip_stage_health_check == true || inputs.run_all_tests == true) && 'true' || 'false' }}
+  SKIP_PR_TEST_HEALTH_CHECK: ${{ (inputs.skip_pr_test_health_check == true || inputs.run_all_tests == true) && 'true' || 'false' }}
   FORCE_REBUILD_DEEPEP: '1'
   PR_TEST_BYPASS_MAINTENANCE_ON_MAIN: ${{ github.ref == 'refs/heads/main' && 'true' || 'false' }}
   USE_VENV: false
@@ -96,7 +96,7 @@ jobs:
       runs_on: x64-kernel-build-node
       job_display_name: Build Wheel
       git_ref: ${{ inputs.git_ref || '' }}
-      skip_stage_health_check: ${{ inputs.skip_stage_health_check == true }}
+      skip_pr_test_health_check: ${{ inputs.skip_pr_test_health_check == true }}
     secrets: inherit
 
   # =============================================== extra-a (1-/2-gpu) ===============================================
diff --git a/.github/workflows/pr-test-jit-kernel.yml b/.github/workflows/pr-test-jit-kernel.yml
index 9e3e597aa..d14c4fc3b 100644
--- a/.github/workflows/pr-test-jit-kernel.yml
+++ b/.github/workflows/pr-test-jit-kernel.yml
@@ -23,7 +23,7 @@ on:
         required: false
         type: string
         default: 'false'
-      skip_stage_health_check:
+      skip_pr_test_health_check:
         required: false
         type: boolean
         default: false
@@ -35,7 +35,7 @@ env:
   SGLANG_CUDA_COREDUMP: "1"
   SGLANG_JIT_DEEPGEMM_FAST_WARMUP: true
   PR_TEST_BYPASS_MAINTENANCE_ON_MAIN: ${{ github.ref == 'refs/heads/main' && 'true' || 'false' }}
-  SKIP_STAGE_HEALTH_CHECK: ${{ inputs.skip_stage_health_check == true && 'true' || 'false' }}
+  SKIP_PR_TEST_HEALTH_CHECK: ${{ inputs.skip_pr_test_health_check == true && 'true' || 'false' }}
 
 jobs:
   jit-kernel-unit-test:
@@ -49,7 +49,7 @@ jobs:
         with:
           ref: ${{ inputs.git_ref || github.sha }}
 
-      - uses: ./.github/actions/check-stage-health
+      - uses: ./.github/actions/check-pr-test-health
 
       - uses: ./.github/actions/check-maintenance
 
@@ -76,7 +76,7 @@ jobs:
         timeout-minutes: 30
         run: |
           cd test/
-          python3 run_suite.py --hw cuda --suite stage-b-kernel-unit-1-gpu-large
+          python3 run_suite.py --hw cuda --suite base-b-kernel-unit-1-gpu-large
 
   jit-kernel-multigpu-unit-test:
     if: |
@@ -114,7 +114,7 @@ jobs:
         timeout-minutes: 45
         run: |
           cd test/
-          python3 run_suite.py --hw cuda --suite stage-b-kernel-unit-8-gpu-h200
+          python3 run_suite.py --hw cuda --suite base-b-kernel-unit-8-gpu-h200
 
   jit-kernel-benchmark-test:
     if: |
@@ -127,7 +127,7 @@ jobs:
         with:
           ref: ${{ inputs.git_ref || github.sha }}
 
-      - uses: ./.github/actions/check-stage-health
+      - uses: ./.github/actions/check-pr-test-health
 
       - uses: ./.github/actions/check-maintenance
 
@@ -154,7 +154,7 @@ jobs:
         timeout-minutes: 45
         run: |
           cd test/
-          python3 run_suite.py --hw cuda --suite stage-b-kernel-benchmark-1-gpu-large
+          python3 run_suite.py --hw cuda --suite base-b-kernel-benchmark-1-gpu-large
 
   jit-kernel-b200-test:
     if: |
@@ -167,7 +167,7 @@ jobs:
         with:
           ref: ${{ inputs.git_ref || github.sha }}
 
-      - uses: ./.github/actions/check-stage-health
+      - uses: ./.github/actions/check-pr-test-health
 
       - uses: ./.github/actions/check-maintenance
 
@@ -194,4 +194,4 @@ jobs:
         timeout-minutes: 30
         run: |
           cd test/
-          python3 run_suite.py --hw cuda --suite stage-b-kernel-unit-1-gpu-b200
+          python3 run_suite.py --hw cuda --suite base-b-kernel-unit-1-gpu-b200
diff --git a/.github/workflows/pr-test-multimodal-gen.yml b/.github/workflows/pr-test-multimodal-gen.yml
index 4ebdb2cd4..b5cf56347 100644
--- a/.github/workflows/pr-test-multimodal-gen.yml
+++ b/.github/workflows/pr-test-multimodal-gen.yml
@@ -31,7 +31,7 @@ on:
         required: false
         type: string
         default: 'false'
-      skip_stage_health_check:
+      skip_pr_test_health_check:
         required: false
         type: string
         default: 'false'
@@ -42,7 +42,7 @@ env:
   SGLANG_IS_IN_CI: true
   SGLANG_CUDA_COREDUMP: "1"
   PR_TEST_BYPASS_MAINTENANCE_ON_MAIN: ${{ github.ref == 'refs/heads/main' && 'true' || 'false' }}
-  SKIP_STAGE_HEALTH_CHECK: ${{ inputs.skip_stage_health_check == 'true' }}
+  SKIP_PR_TEST_HEALTH_CHECK: ${{ inputs.skip_pr_test_health_check == 'true' }}
 
 jobs:
   compute-diffusion-partitions:
@@ -90,7 +90,7 @@ jobs:
         with:
           ref: ${{ inputs.git_ref || github.sha }}
 
-      - uses: ./.github/actions/check-stage-health
+      - uses: ./.github/actions/check-pr-test-health
 
       - uses: ./.github/actions/check-maintenance
 
@@ -163,7 +163,7 @@ jobs:
         with:
           ref: ${{ inputs.git_ref || github.sha }}
 
-      - uses: ./.github/actions/check-stage-health
+      - uses: ./.github/actions/check-pr-test-health
 
       - uses: ./.github/actions/check-maintenance
 
@@ -230,7 +230,7 @@ jobs:
         with:
           ref: ${{ inputs.git_ref || github.sha }}
 
-      - uses: ./.github/actions/check-stage-health
+      - uses: ./.github/actions/check-pr-test-health
 
       - uses: ./.github/actions/check-maintenance
 
@@ -275,7 +275,7 @@ jobs:
         with:
           ref: ${{ inputs.git_ref || github.sha }}
 
-      - uses: ./.github/actions/check-stage-health
+      - uses: ./.github/actions/check-pr-test-health
 
       - uses: ./.github/actions/check-maintenance
 
@@ -327,7 +327,7 @@ jobs:
         with:
           ref: ${{ inputs.git_ref || github.sha }}
 
-      - uses: ./.github/actions/check-stage-health
+      - uses: ./.github/actions/check-pr-test-health
 
       - uses: ./.github/actions/check-maintenance
 
diff --git a/.github/workflows/pr-test-sgl-kernel.yml b/.github/workflows/pr-test-sgl-kernel.yml
index af1552954..272b0951f 100644
--- a/.github/workflows/pr-test-sgl-kernel.yml
+++ b/.github/workflows/pr-test-sgl-kernel.yml
@@ -16,7 +16,7 @@ on:
         required: false
         type: string
         default: ''
-      skip_stage_health_check:
+      skip_pr_test_health_check:
         required: false
         type: boolean
         default: false
@@ -27,7 +27,7 @@ env:
   SGLANG_IS_IN_CI: true
   SGLANG_CUDA_COREDUMP: "1"
   PR_TEST_BYPASS_MAINTENANCE_ON_MAIN: ${{ github.ref == 'refs/heads/main' && 'true' || 'false' }}
-  SKIP_STAGE_HEALTH_CHECK: ${{ inputs.skip_stage_health_check == true && 'true' || 'false' }}
+  SKIP_PR_TEST_HEALTH_CHECK: ${{ inputs.skip_pr_test_health_check == true && 'true' || 'false' }}
 
 jobs:
   sgl-kernel-unit-test:
@@ -38,7 +38,7 @@ jobs:
         with:
           ref: ${{ inputs.git_ref || github.sha }}
 
-      - uses: ./.github/actions/check-stage-health
+      - uses: ./.github/actions/check-pr-test-health
 
       - uses: ./.github/actions/check-maintenance
 
@@ -73,7 +73,7 @@ jobs:
         with:
           ref: ${{ inputs.git_ref || github.sha }}
 
-      - uses: ./.github/actions/check-stage-health
+      - uses: ./.github/actions/check-pr-test-health
 
       - uses: ./.github/actions/check-maintenance
 
@@ -120,7 +120,7 @@ jobs:
         with:
           ref: ${{ inputs.git_ref || github.sha }}
 
-      - uses: ./.github/actions/check-stage-health
+      - uses: ./.github/actions/check-pr-test-health
 
       - uses: ./.github/actions/check-maintenance
 
diff --git a/.github/workflows/pr-test-xeon.yml b/.github/workflows/pr-test-xeon.yml
index abaa9401d..3c7ae2bf4 100644
--- a/.github/workflows/pr-test-xeon.yml
+++ b/.github/workflows/pr-test-xeon.yml
@@ -115,7 +115,7 @@ jobs:
         timeout-minutes: 36
         run: |
           docker exec -w /sglang-checkout/ ci_sglang_xeon \
-            bash -c "source /opt/.venv/bin/activate && cd ./test && python3 run_suite.py --hw cpu --suite stage-b-test-cpu"
+            bash -c "source /opt/.venv/bin/activate && cd ./test && python3 run_suite.py --hw cpu --suite base-b-test-cpu"
 
       - name: Change permission
         timeout-minutes: 2
diff --git a/.github/workflows/pr-test.yml b/.github/workflows/pr-test.yml
index d5f4d824c..0d9b81712 100644
--- a/.github/workflows/pr-test.yml
+++ b/.github/workflows/pr-test.yml
@@ -1,4 +1,4 @@
-name: PR Test
+name: PR Test Base
 
 on:
   schedule:
@@ -34,8 +34,8 @@ on:
         required: false
         type: boolean
         default: false
-      skip_stage_health_check:
-        description: "Skip stage health check fast-fail (e.g. for release branch cuts)"
+      skip_pr_test_health_check:
+        description: "Skip PR test health check fast-fail (e.g. for release branch cuts)"
         required: false
         type: boolean
         default: false
@@ -52,7 +52,7 @@ env:
   SGLANG_IS_IN_CI: true
   SGLANG_CUDA_COREDUMP: "1"
   SGLANG_JIT_DEEPGEMM_FAST_WARMUP: true
-  SKIP_STAGE_HEALTH_CHECK: ${{ (inputs.skip_stage_health_check == true || inputs.test_parallel_dispatch == true || inputs.run_all_tests == true) && 'true' || 'false' }}
+  SKIP_PR_TEST_HEALTH_CHECK: ${{ (inputs.skip_pr_test_health_check == true || inputs.test_parallel_dispatch == true || inputs.run_all_tests == true) && 'true' || 'false' }}
   # TEMP: rebuild deepep against the new torch for torch-211-merge PR only — revert before merging to main.
   FORCE_REBUILD_DEEPEP: '1'
   # Schedule / main-branch dispatch / workflow_call from main use refs/heads/main; PR events use refs/pull/*/merge
@@ -89,7 +89,7 @@ jobs:
   # For PRs with the `bypass-fastfail` label: wait jobs run but return success immediately
   # (handled inside the wait-for-jobs action), so downstream stages dispatch in parallel.
 
-  wait-for-stage-a:
+  wait-for-base-a:
     needs: [check-changes, call-gate]
     if: |
       always() &&
@@ -100,7 +100,7 @@ jobs:
       (needs.call-gate.result == 'success' || needs.call-gate.result == 'skipped')
     runs-on: ubuntu-latest
     outputs:
-      stage_a_result: ${{ steps.wait.outputs.result }}
+      base_a_result: ${{ steps.wait.outputs.result }}
     steps:
       - uses: actions/checkout@v4
 
@@ -109,27 +109,27 @@ jobs:
       - uses: ./.github/actions/wait-for-jobs
         id: wait
         with:
-          stage-name: stage-a
+          stage-name: base-a
           jobs: |
             [
-              {"prefix": "stage-a-test-1-gpu-small", "expected_count": ${{ fromJson(needs.check-changes.outputs.partitions)['stage-a-test-1-gpu-small'].size }}},
-              {"prefix": "stage-a-test-cpu", "expected_count": ${{ fromJson(needs.check-changes.outputs.partitions)['stage-a-test-cpu'].size }}}
+              {"prefix": "base-a-test-1-gpu-small", "expected_count": ${{ fromJson(needs.check-changes.outputs.partitions)['base-a-test-1-gpu-small'].size }}},
+              {"prefix": "base-a-test-cpu", "expected_count": ${{ fromJson(needs.check-changes.outputs.partitions)['base-a-test-cpu'].size }}}
             ]
           max-wait-minutes: '240'
 
-  wait-for-stage-b:
-    needs: [check-changes, call-gate, wait-for-stage-a]
+  wait-for-base-b:
+    needs: [check-changes, call-gate, wait-for-base-a]
     if: |
       always() &&
       !cancelled() &&
       github.event_name == 'pull_request' &&
       inputs.test_parallel_dispatch != true &&
       (needs.check-changes.outputs.main_package == 'true' || needs.check-changes.outputs.sgl_kernel == 'true') &&
-      (needs.wait-for-stage-a.result == 'success' || needs.wait-for-stage-a.result == 'skipped') &&
+      (needs.wait-for-base-a.result == 'success' || needs.wait-for-base-a.result == 'skipped') &&
       (needs.call-gate.result == 'success' || needs.call-gate.result == 'skipped')
     runs-on: ubuntu-latest
     outputs:
-      stage_b_result: ${{ steps.wait.outputs.result }}
+      base_b_result: ${{ steps.wait.outputs.result }}
     steps:
       - uses: actions/checkout@v4
 
@@ -138,13 +138,13 @@ jobs:
       - uses: ./.github/actions/wait-for-jobs
         id: wait
         with:
-          stage-name: stage-b
+          stage-name: base-b
           jobs: |
             [
-              {"prefix": "stage-b-test-1-gpu-small", "expected_count": ${{ fromJson(needs.check-changes.outputs.partitions)['stage-b-test-1-gpu-small'].size }}},
-              {"prefix": "stage-b-test-1-gpu-large", "expected_count": ${{ fromJson(needs.check-changes.outputs.partitions)['stage-b-test-1-gpu-large'].size }}},
-              {"prefix": "stage-b-test-2-gpu-large", "expected_count": ${{ fromJson(needs.check-changes.outputs.partitions)['stage-b-test-2-gpu-large'].size }}},
-              {"prefix": "stage-b-test-4-gpu-b200", "expected_count": ${{ fromJson(needs.check-changes.outputs.partitions)['stage-b-test-4-gpu-b200'].size }}}
+              {"prefix": "base-b-test-1-gpu-small", "expected_count": ${{ fromJson(needs.check-changes.outputs.partitions)['base-b-test-1-gpu-small'].size }}},
+              {"prefix": "base-b-test-1-gpu-large", "expected_count": ${{ fromJson(needs.check-changes.outputs.partitions)['base-b-test-1-gpu-large'].size }}},
+              {"prefix": "base-b-test-2-gpu-large", "expected_count": ${{ fromJson(needs.check-changes.outputs.partitions)['base-b-test-2-gpu-large'].size }}},
+              {"prefix": "base-b-test-4-gpu-b200", "expected_count": ${{ fromJson(needs.check-changes.outputs.partitions)['base-b-test-4-gpu-b200'].size }}}
             ]
           max-wait-minutes: '480'
 
@@ -180,7 +180,7 @@ jobs:
       runs_on: x64-kernel-build-node
       job_display_name: Build Wheel
       git_ref: ${{ inputs.git_ref || '' }}
-      skip_stage_health_check: ${{ inputs.skip_stage_health_check == true }}
+      skip_pr_test_health_check: ${{ inputs.skip_pr_test_health_check == true }}
     secrets: inherit
 
   sgl-kernel-build-wheels-arm:
@@ -198,7 +198,7 @@ jobs:
       job_display_name: Build Wheel Arm
       arch_suffix: '-aarch64'
       git_ref: ${{ inputs.git_ref || '' }}
-      skip_stage_health_check: ${{ inputs.skip_stage_health_check == true }}
+      skip_pr_test_health_check: ${{ inputs.skip_pr_test_health_check == true }}
     secrets: inherit
 
   call-sgl-kernel-tests:
@@ -213,7 +213,7 @@ jobs:
       runs_on_map: ${{ needs.check-changes.outputs.runs_on_map }}
       sgl_kernel: ${{ needs.check-changes.outputs.sgl_kernel }}
       git_ref: ${{ inputs.git_ref || '' }}
-      skip_stage_health_check: ${{ inputs.skip_stage_health_check == true }}
+      skip_pr_test_health_check: ${{ inputs.skip_pr_test_health_check == true }}
     secrets: inherit
 
   # =============================================== jit-kernel ====================================================
@@ -234,18 +234,18 @@ jobs:
       sgl_kernel: ${{ needs.check-changes.outputs.sgl_kernel }}
       git_ref: ${{ inputs.git_ref || '' }}
       test_parallel_dispatch: ${{ inputs.test_parallel_dispatch == true && 'true' || 'false' }}
-      skip_stage_health_check: ${{ inputs.skip_stage_health_check == true }}
+      skip_pr_test_health_check: ${{ inputs.skip_pr_test_health_check == true }}
     secrets: inherit
 
   # =============================================== primary ====================================================
 
   # Runs on 5090 (32GB, SM120)
-  stage-a-test-1-gpu-small:
+  base-a-test-1-gpu-small:
     needs: [check-changes, call-gate, sgl-kernel-build-wheels]
     if: ${{ !failure() && !cancelled() }}
     uses: ./.github/workflows/_pr-test-stage.yml
     with:
-      self_name: stage-a-test-1-gpu-small
+      self_name: base-a-test-1-gpu-small
       runner_config: 1-gpu-small
       check_changes: ${{ toJson(needs.check-changes.outputs) }}
       caller_inputs: ${{ toJson(inputs) }}
@@ -253,7 +253,7 @@ jobs:
       run_timeout_minutes: '10'
     secrets: inherit
 
-  stage-a-test-cpu:
+  base-a-test-cpu:
     needs: [check-changes, call-gate]
     if: |
       always() &&
@@ -263,9 +263,9 @@ jobs:
     timeout-minutes: 240
     strategy:
       fail-fast: false
-      max-parallel: ${{ fromJson(needs.check-changes.outputs.partitions)['stage-a-test-cpu'].max_parallel }}
+      max-parallel: ${{ fromJson(needs.check-changes.outputs.partitions)['base-a-test-cpu'].max_parallel }}
       matrix:
-        partition: ${{ fromJson(needs.check-changes.outputs.partitions)['stage-a-test-cpu'].arr }}
+        partition: ${{ fromJson(needs.check-changes.outputs.partitions)['base-a-test-cpu'].arr }}
     steps:
       - name: Free disk space
         run: |
@@ -277,7 +277,7 @@ jobs:
         with:
           ref: ${{ inputs.git_ref || github.sha }}
 
-      - uses: ./.github/actions/check-stage-health
+      - uses: ./.github/actions/check-pr-test-health
 
       - uses: ./.github/actions/check-maintenance
 
@@ -316,15 +316,15 @@ jobs:
           CONTINUE_ON_ERROR_FLAG: ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}
         run: |
           cd test/
-          python3 run_suite.py --hw cpu --suite stage-a-test-cpu --auto-partition-id ${{ matrix.partition }} --auto-partition-size ${{ fromJson(needs.check-changes.outputs.partitions)['stage-a-test-cpu'].size }} $CONTINUE_ON_ERROR_FLAG
+          python3 run_suite.py --hw cpu --suite base-a-test-cpu --auto-partition-id ${{ matrix.partition }} --auto-partition-size ${{ fromJson(needs.check-changes.outputs.partitions)['base-a-test-cpu'].size }} $CONTINUE_ON_ERROR_FLAG
 
   # Runs on 5090 (32GB, SM120)
-  stage-b-test-1-gpu-small:
-    needs: [check-changes, call-gate, wait-for-stage-a, sgl-kernel-build-wheels]
+  base-b-test-1-gpu-small:
+    needs: [check-changes, call-gate, wait-for-base-a, sgl-kernel-build-wheels]
     if: ${{ !failure() && !cancelled() }}
     uses: ./.github/workflows/_pr-test-stage.yml
     with:
-      self_name: stage-b-test-1-gpu-small
+      self_name: base-b-test-1-gpu-small
       runner_config: 1-gpu-small
       check_changes: ${{ toJson(needs.check-changes.outputs) }}
       caller_inputs: ${{ toJson(inputs) }}
@@ -333,12 +333,12 @@ jobs:
     secrets: inherit
 
   # Runs on H100 (80GB, SM90) - tests that don't pass on 5090 (FA3, FP8, high VRAM, etc.)
-  stage-b-test-1-gpu-large:
-    needs: [check-changes, call-gate, wait-for-stage-a, sgl-kernel-build-wheels]
+  base-b-test-1-gpu-large:
+    needs: [check-changes, call-gate, wait-for-base-a, sgl-kernel-build-wheels]
     if: ${{ !failure() && !cancelled() }}
     uses: ./.github/workflows/_pr-test-stage.yml
     with:
-      self_name: stage-b-test-1-gpu-large
+      self_name: base-b-test-1-gpu-large
       runner_config: 1-gpu-large
       check_changes: ${{ toJson(needs.check-changes.outputs) }}
       caller_inputs: ${{ toJson(inputs) }}
@@ -347,12 +347,12 @@ jobs:
       timeout_per_file: '1800'
     secrets: inherit
 
-  stage-b-test-2-gpu-large:
-    needs: [check-changes, call-gate, wait-for-stage-a, sgl-kernel-build-wheels]
+  base-b-test-2-gpu-large:
+    needs: [check-changes, call-gate, wait-for-base-a, sgl-kernel-build-wheels]
     if: ${{ !failure() && !cancelled() }}
     uses: ./.github/workflows/_pr-test-stage.yml
     with:
-      self_name: stage-b-test-2-gpu-large
+      self_name: base-b-test-2-gpu-large
       runner_config: 2-gpu-large
       check_changes: ${{ toJson(needs.check-changes.outputs) }}
       caller_inputs: ${{ toJson(inputs) }}
@@ -360,12 +360,12 @@ jobs:
       run_timeout_minutes: '30'
     secrets: inherit
 
-  stage-b-test-4-gpu-b200:
-    needs: [check-changes, call-gate, wait-for-stage-a, sgl-kernel-build-wheels]
+  base-b-test-4-gpu-b200:
+    needs: [check-changes, call-gate, wait-for-base-a, sgl-kernel-build-wheels]
     if: ${{ !failure() && !cancelled() }}
     uses: ./.github/workflows/_pr-test-stage.yml
     with:
-      self_name: stage-b-test-4-gpu-b200
+      self_name: base-b-test-4-gpu-b200
       runner_config: 4-gpu-b200
       check_changes: ${{ toJson(needs.check-changes.outputs) }}
       caller_inputs: ${{ toJson(inputs) }}
@@ -391,15 +391,15 @@ jobs:
       git_ref: ${{ inputs.git_ref || '' }}
       test_parallel_dispatch: ${{ inputs.test_parallel_dispatch == true && 'true' || 'false' }}
       caller_needs_failure: ${{ (needs.call-gate.result == 'failure' || needs.sgl-kernel-build-wheels.result == 'failure' || needs.check-changes.result == 'failure') && 'true' || 'false' }}
-      skip_stage_health_check: ${{ inputs.skip_stage_health_check == true && 'true' || 'false' }}
+      skip_pr_test_health_check: ${{ inputs.skip_pr_test_health_check == true && 'true' || 'false' }}
     secrets: inherit
 
-  stage-c-test-4-gpu-h100:
-    needs: [check-changes, call-gate, wait-for-stage-b, sgl-kernel-build-wheels]
+  base-c-test-4-gpu-h100:
+    needs: [check-changes, call-gate, wait-for-base-b, sgl-kernel-build-wheels]
     if: ${{ !failure() && !cancelled() }}
     uses: ./.github/workflows/_pr-test-stage.yml
     with:
-      self_name: stage-c-test-4-gpu-h100
+      self_name: base-c-test-4-gpu-h100
       runner_config: 4-gpu-h100
       check_changes: ${{ toJson(needs.check-changes.outputs) }}
       caller_inputs: ${{ toJson(inputs) }}
@@ -407,12 +407,12 @@ jobs:
       run_timeout_minutes: '30'
     secrets: inherit
 
-  stage-c-test-8-gpu-h200:
-    needs: [check-changes, call-gate, wait-for-stage-b, sgl-kernel-build-wheels]
+  base-c-test-8-gpu-h200:
+    needs: [check-changes, call-gate, wait-for-base-b, sgl-kernel-build-wheels]
     if: ${{ !failure() && !cancelled() }}
     uses: ./.github/workflows/_pr-test-stage.yml
     with:
-      self_name: stage-c-test-8-gpu-h200
+      self_name: base-c-test-8-gpu-h200
       runner_config: 8-gpu-h200
       check_changes: ${{ toJson(needs.check-changes.outputs) }}
       caller_inputs: ${{ toJson(inputs) }}
@@ -427,12 +427,12 @@ jobs:
       warmup_timeout_minutes: '60'
     secrets: inherit
 
-  stage-c-test-8-gpu-h20:
-    needs: [check-changes, call-gate, wait-for-stage-b, sgl-kernel-build-wheels]
+  base-c-test-8-gpu-h20:
+    needs: [check-changes, call-gate, wait-for-base-b, sgl-kernel-build-wheels]
     if: ${{ !failure() && !cancelled() }}
     uses: ./.github/workflows/_pr-test-stage.yml
     with:
-      self_name: stage-c-test-8-gpu-h20
+      self_name: base-c-test-8-gpu-h20
       runner_config: 8-gpu-h20
       check_changes: ${{ toJson(needs.check-changes.outputs) }}
       caller_inputs: ${{ toJson(inputs) }}
@@ -440,12 +440,12 @@ jobs:
       run_timeout_minutes: '30'
     secrets: inherit
 
-  stage-c-test-deepep-4-gpu-h100:
-    needs: [check-changes, call-gate, wait-for-stage-b, sgl-kernel-build-wheels]
+  base-c-test-deepep-4-gpu-h100:
+    needs: [check-changes, call-gate, wait-for-base-b, sgl-kernel-build-wheels]
     if: ${{ !failure() && !cancelled() }}
     uses: ./.github/workflows/_pr-test-stage.yml
     with:
-      self_name: stage-c-test-deepep-4-gpu-h100
+      self_name: base-c-test-deepep-4-gpu-h100
       runner_config: deepep-4-gpu-h100
       check_changes: ${{ toJson(needs.check-changes.outputs) }}
       caller_inputs: ${{ toJson(inputs) }}
@@ -455,12 +455,12 @@ jobs:
       warmup_server_models: 'lmsys/sglang-ci-dsv3-test:4'
     secrets: inherit
 
-  stage-c-test-4-gpu-b200:
-    needs: [check-changes, call-gate, wait-for-stage-b, sgl-kernel-build-wheels]
+  base-c-test-4-gpu-b200:
+    needs: [check-changes, call-gate, wait-for-base-b, sgl-kernel-build-wheels]
     if: ${{ !failure() && !cancelled() }}
     uses: ./.github/workflows/_pr-test-stage.yml
     with:
-      self_name: stage-c-test-4-gpu-b200
+      self_name: base-c-test-4-gpu-b200
       runner_config: 4-gpu-b200
       check_changes: ${{ toJson(needs.check-changes.outputs) }}
       caller_inputs: ${{ toJson(inputs) }}
@@ -469,12 +469,12 @@ jobs:
       timeout_per_file: '1800'
     secrets: inherit
 
-  stage-c-test-dsv4-4-gpu-b200:
-    needs: [check-changes, call-gate, wait-for-stage-b, sgl-kernel-build-wheels]
+  base-c-test-dsv4-4-gpu-b200:
+    needs: [check-changes, call-gate, wait-for-base-b, sgl-kernel-build-wheels]
     if: ${{ !failure() && !cancelled() }}
     uses: ./.github/workflows/_pr-test-stage.yml
     with:
-      self_name: stage-c-test-dsv4-4-gpu-b200
+      self_name: base-c-test-dsv4-4-gpu-b200
       runner_config: dsv4-4-gpu-b200
       check_changes: ${{ toJson(needs.check-changes.outputs) }}
       caller_inputs: ${{ toJson(inputs) }}
@@ -483,12 +483,12 @@ jobs:
       timeout_per_file: '1800'
     secrets: inherit
 
-  stage-c-test-dsv4-8-gpu-h200:
-    needs: [check-changes, call-gate, wait-for-stage-b, sgl-kernel-build-wheels]
+  base-c-test-dsv4-8-gpu-h200:
+    needs: [check-changes, call-gate, wait-for-base-b, sgl-kernel-build-wheels]
     if: ${{ !failure() && !cancelled() }}
     uses: ./.github/workflows/_pr-test-stage.yml
     with:
-      self_name: stage-c-test-dsv4-8-gpu-h200
+      self_name: base-c-test-dsv4-8-gpu-h200
       runner_config: dsv4-8-gpu-h200
       check_changes: ${{ toJson(needs.check-changes.outputs) }}
       caller_inputs: ${{ toJson(inputs) }}
@@ -507,27 +507,27 @@ jobs:
         sgl-kernel-build-wheels-arm,
         call-sgl-kernel-tests,
 
-        wait-for-stage-a,
-        wait-for-stage-b,
+        wait-for-base-a,
+        wait-for-base-b,
 
         call-jit-kernel-tests,
 
         call-multimodal-gen-tests,
 
-        stage-a-test-1-gpu-small,
-        stage-a-test-cpu,
-        stage-b-test-1-gpu-small,
-        stage-b-test-1-gpu-large,
-        stage-b-test-2-gpu-large,
-        stage-b-test-4-gpu-b200,
-        stage-c-test-4-gpu-h100,
-        stage-c-test-8-gpu-h20,
-        stage-c-test-8-gpu-h200,
-        stage-c-test-deepep-4-gpu-h100,
-        stage-c-test-4-gpu-b200,
-        stage-c-test-dsv4-4-gpu-b200,
-        stage-c-test-dsv4-8-gpu-h200,
-        # stage-c-test-4-gpu-gb200,  # Temporarily disabled — no GB200 runner
+        base-a-test-1-gpu-small,
+        base-a-test-cpu,
+        base-b-test-1-gpu-small,
+        base-b-test-1-gpu-large,
+        base-b-test-2-gpu-large,
+        base-b-test-4-gpu-b200,
+        base-c-test-4-gpu-h100,
+        base-c-test-8-gpu-h20,
+        base-c-test-8-gpu-h200,
+        base-c-test-deepep-4-gpu-h100,
+        base-c-test-4-gpu-b200,
+        base-c-test-dsv4-4-gpu-b200,
+        base-c-test-dsv4-8-gpu-h200,
+        # base-c-test-4-gpu-gb200,  # Temporarily disabled — no GB200 runner
       ]
     if: always()
     runs-on: ubuntu-latest
diff --git a/.github/workflows/release-branch-cut.yml b/.github/workflows/release-branch-cut.yml
index a4ed645d5..2afd56043 100644
--- a/.github/workflows/release-branch-cut.yml
+++ b/.github/workflows/release-branch-cut.yml
@@ -151,7 +151,7 @@ jobs:
     with:
       git_ref: ${{ needs.cut-release-branch.outputs.branch_name }}
       run_all_tests: true
-      skip_stage_health_check: true
+      skip_pr_test_health_check: true
     secrets: inherit
 
   run-pr-tests-amd:
diff --git a/python/sglang/jit_kernel/benchmark/bench_activation.py b/python/sglang/jit_kernel/benchmark/bench_activation.py
index 3f0ba2f6c..1144026b9 100644
--- a/python/sglang/jit_kernel/benchmark/bench_activation.py
+++ b/python/sglang/jit_kernel/benchmark/bench_activation.py
@@ -19,7 +19,7 @@ from sglang.jit_kernel.benchmark.utils import (
 )
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=30, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=30, suite="base-b-kernel-benchmark-1-gpu-large")
 
 
 @torch.compile
diff --git a/python/sglang/jit_kernel/benchmark/bench_awq_dequantize.py b/python/sglang/jit_kernel/benchmark/bench_awq_dequantize.py
index 81422a597..85640b14c 100644
--- a/python/sglang/jit_kernel/benchmark/bench_awq_dequantize.py
+++ b/python/sglang/jit_kernel/benchmark/bench_awq_dequantize.py
@@ -9,7 +9,7 @@ from sglang.jit_kernel.benchmark.utils import run_benchmark
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.utils import is_in_ci
 
-register_cuda_ci(est_time=5, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=5, suite="base-b-kernel-benchmark-1-gpu-large")
 
 try:
     from sgl_kernel import awq_dequantize as aot_awq_dequantize
diff --git a/python/sglang/jit_kernel/benchmark/bench_cast.py b/python/sglang/jit_kernel/benchmark/bench_cast.py
index 97c71bcb0..4be874ce5 100644
--- a/python/sglang/jit_kernel/benchmark/bench_cast.py
+++ b/python/sglang/jit_kernel/benchmark/bench_cast.py
@@ -10,7 +10,7 @@ from sglang.jit_kernel.benchmark.utils import (
 from sglang.jit_kernel.cast import downcast_fp8 as downcast_fp8_jit
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=10, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=10, suite="base-b-kernel-benchmark-1-gpu-large")
 
 DEVICE = DEFAULT_DEVICE
 DTYPE = torch.bfloat16
diff --git a/python/sglang/jit_kernel/benchmark/bench_clamp_position.py b/python/sglang/jit_kernel/benchmark/bench_clamp_position.py
index 8b33edc6b..dfbb5b1e4 100644
--- a/python/sglang/jit_kernel/benchmark/bench_clamp_position.py
+++ b/python/sglang/jit_kernel/benchmark/bench_clamp_position.py
@@ -13,7 +13,7 @@ from sglang.jit_kernel.clamp_position import clamp_position_cuda
 from sglang.srt.utils import get_compiler_backend
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=13, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=13, suite="base-b-kernel-benchmark-1-gpu-large")
 register_amd_ci(est_time=16, suite="jit-kernel-unit-test-amd")
 
 SIZE_LIST = get_benchmark_range(
diff --git a/python/sglang/jit_kernel/benchmark/bench_concat_mla.py b/python/sglang/jit_kernel/benchmark/bench_concat_mla.py
index 8129b7db1..0a0d98aa3 100644
--- a/python/sglang/jit_kernel/benchmark/bench_concat_mla.py
+++ b/python/sglang/jit_kernel/benchmark/bench_concat_mla.py
@@ -12,7 +12,7 @@ from sglang.jit_kernel.concat_mla import concat_mla_k as jit_k
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.utils import is_in_ci
 
-register_cuda_ci(est_time=6, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=6, suite="base-b-kernel-benchmark-1-gpu-large")
 
 IS_CI = is_in_ci()
 
diff --git a/python/sglang/jit_kernel/benchmark/bench_custom_all_reduce.py b/python/sglang/jit_kernel/benchmark/bench_custom_all_reduce.py
index 4f36f1a48..79167e53b 100644
--- a/python/sglang/jit_kernel/benchmark/bench_custom_all_reduce.py
+++ b/python/sglang/jit_kernel/benchmark/bench_custom_all_reduce.py
@@ -26,7 +26,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(
     est_time=120,
-    suite="stage-b-kernel-benchmark-1-gpu-large",
+    suite="base-b-kernel-benchmark-1-gpu-large",
     disabled="requires multi-GPU, self-skips in CI",
 )
 
diff --git a/python/sglang/jit_kernel/benchmark/bench_fused_qknorm_rope.py b/python/sglang/jit_kernel/benchmark/bench_fused_qknorm_rope.py
index 905d41d5d..5ae05f78b 100644
--- a/python/sglang/jit_kernel/benchmark/bench_fused_qknorm_rope.py
+++ b/python/sglang/jit_kernel/benchmark/bench_fused_qknorm_rope.py
@@ -20,7 +20,7 @@ from sglang.jit_kernel.fused_qknorm_rope import (
 )
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=6, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=6, suite="base-b-kernel-benchmark-1-gpu-large")
 
 try:
     from sgl_kernel import fused_qk_norm_rope as fused_qk_norm_rope_aot
diff --git a/python/sglang/jit_kernel/benchmark/bench_hadamard.py b/python/sglang/jit_kernel/benchmark/bench_hadamard.py
index 3da9ec484..704107c2d 100644
--- a/python/sglang/jit_kernel/benchmark/bench_hadamard.py
+++ b/python/sglang/jit_kernel/benchmark/bench_hadamard.py
@@ -16,7 +16,7 @@ from sglang.jit_kernel.benchmark.utils import (
 from sglang.jit_kernel.hadamard import hadamard_transform
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=5, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=5, suite="base-b-kernel-benchmark-1-gpu-large")
 
 # AOT kernel: might not be available in all environments.
 # This is used for performance baseline comparison.
diff --git a/python/sglang/jit_kernel/benchmark/bench_hicache.py b/python/sglang/jit_kernel/benchmark/bench_hicache.py
index d0e4a31aa..ba897fc12 100644
--- a/python/sglang/jit_kernel/benchmark/bench_hicache.py
+++ b/python/sglang/jit_kernel/benchmark/bench_hicache.py
@@ -33,7 +33,7 @@ from sglang.jit_kernel.hicache import (
 )
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=29, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=29, suite="base-b-kernel-benchmark-1-gpu-large")
 
 DISABLE_TORCH = os.environ.get("DISABLE_TORCH", "0") == "1"
 PAGE_SIZE = 1
diff --git a/python/sglang/jit_kernel/benchmark/bench_hisparse.py b/python/sglang/jit_kernel/benchmark/bench_hisparse.py
index 6eb5fb5f0..2fe0ef901 100644
--- a/python/sglang/jit_kernel/benchmark/bench_hisparse.py
+++ b/python/sglang/jit_kernel/benchmark/bench_hisparse.py
@@ -9,7 +9,7 @@ from sglang.jit_kernel.benchmark.utils import DEFAULT_DEVICE, DEFAULT_DTYPE
 from sglang.jit_kernel.hisparse import load_cache_to_device_buffer_mla
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=12, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=12, suite="base-b-kernel-benchmark-1-gpu-large")
 
 DEVICE = DEFAULT_DEVICE
 DTYPE = DEFAULT_DTYPE
diff --git a/python/sglang/jit_kernel/benchmark/bench_mla_kv_pack_quantize_fp8.py b/python/sglang/jit_kernel/benchmark/bench_mla_kv_pack_quantize_fp8.py
index a797e80ba..6d0f60b1c 100644
--- a/python/sglang/jit_kernel/benchmark/bench_mla_kv_pack_quantize_fp8.py
+++ b/python/sglang/jit_kernel/benchmark/bench_mla_kv_pack_quantize_fp8.py
@@ -20,7 +20,7 @@ from sglang.jit_kernel.mla_kv_pack_quantize_fp8 import (
 from sglang.jit_kernel.utils import is_arch_support_pdl
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=15, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=15, suite="base-b-kernel-benchmark-1-gpu-large")
 
 
 @triton.jit
diff --git a/python/sglang/jit_kernel/benchmark/bench_mxfp8_moe.py b/python/sglang/jit_kernel/benchmark/bench_mxfp8_moe.py
index 11ae67904..a4215cc1b 100644
--- a/python/sglang/jit_kernel/benchmark/bench_mxfp8_moe.py
+++ b/python/sglang/jit_kernel/benchmark/bench_mxfp8_moe.py
@@ -13,7 +13,7 @@ from sglang.jit_kernel.mxfp8 import (
 )
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=5, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=5, suite="base-b-kernel-benchmark-1-gpu-large")
 
 
 def is_sm100_supported(device=None) -> bool:
diff --git a/python/sglang/jit_kernel/benchmark/bench_norm.py b/python/sglang/jit_kernel/benchmark/bench_norm.py
index 345388ef7..6adc7ac64 100644
--- a/python/sglang/jit_kernel/benchmark/bench_norm.py
+++ b/python/sglang/jit_kernel/benchmark/bench_norm.py
@@ -11,7 +11,7 @@ from sglang.jit_kernel.norm import fused_add_rmsnorm as jit_fused_add_rmsnorm
 from sglang.jit_kernel.norm import rmsnorm as jit_rmsnorm
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=30, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=30, suite="base-b-kernel-benchmark-1-gpu-large")
 
 
 DTYPE = torch.bfloat16
diff --git a/python/sglang/jit_kernel/benchmark/bench_nvfp4_blockwise_moe.py b/python/sglang/jit_kernel/benchmark/bench_nvfp4_blockwise_moe.py
index 98da03e72..0a4a1431e 100644
--- a/python/sglang/jit_kernel/benchmark/bench_nvfp4_blockwise_moe.py
+++ b/python/sglang/jit_kernel/benchmark/bench_nvfp4_blockwise_moe.py
@@ -15,7 +15,7 @@ from sglang.jit_kernel.nvfp4 import (
 from sglang.srt.utils import is_sm100_supported
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=5, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=5, suite="base-b-kernel-benchmark-1-gpu-large")
 
 FLOAT4_E2M1_MAX = 6.0
 FLOAT8_E4M3_MAX = torch.finfo(torch.float8_e4m3fn).max
diff --git a/python/sglang/jit_kernel/benchmark/bench_nvfp4_quant.py b/python/sglang/jit_kernel/benchmark/bench_nvfp4_quant.py
index 2be07d3d6..5d516a9ee 100644
--- a/python/sglang/jit_kernel/benchmark/bench_nvfp4_quant.py
+++ b/python/sglang/jit_kernel/benchmark/bench_nvfp4_quant.py
@@ -10,7 +10,7 @@ from sglang.jit_kernel.nvfp4 import scaled_fp4_quant
 from sglang.srt.utils import is_sm100_supported
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=5, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=5, suite="base-b-kernel-benchmark-1-gpu-large")
 
 FLOAT4_E2M1_MAX = 6.0
 FLOAT8_E4M3_MAX = torch.finfo(torch.float8_e4m3fn).max
diff --git a/python/sglang/jit_kernel/benchmark/bench_nvfp4_scaled_mm.py b/python/sglang/jit_kernel/benchmark/bench_nvfp4_scaled_mm.py
index 4278f0348..251840fe6 100644
--- a/python/sglang/jit_kernel/benchmark/bench_nvfp4_scaled_mm.py
+++ b/python/sglang/jit_kernel/benchmark/bench_nvfp4_scaled_mm.py
@@ -10,7 +10,7 @@ from sglang.jit_kernel.nvfp4 import cutlass_scaled_fp4_mm, scaled_fp4_quant
 from sglang.srt.utils import is_sm100_supported, is_sm120_supported
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=5, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=5, suite="base-b-kernel-benchmark-1-gpu-large")
 
 FLOAT4_E2M1_MAX = 6.0
 FLOAT8_E4M3_MAX = torch.finfo(torch.float8_e4m3fn).max
diff --git a/python/sglang/jit_kernel/benchmark/bench_per_tensor_quant_fp8.py b/python/sglang/jit_kernel/benchmark/bench_per_tensor_quant_fp8.py
index a061639b8..e5c9f486b 100644
--- a/python/sglang/jit_kernel/benchmark/bench_per_tensor_quant_fp8.py
+++ b/python/sglang/jit_kernel/benchmark/bench_per_tensor_quant_fp8.py
@@ -8,7 +8,7 @@ from sglang.jit_kernel.benchmark.utils import get_benchmark_range, run_benchmark
 from sglang.jit_kernel.per_tensor_quant_fp8 import per_tensor_quant_fp8
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=5, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=5, suite="base-b-kernel-benchmark-1-gpu-large")
 
 try:
     from vllm import _custom_ops as ops
diff --git a/python/sglang/jit_kernel/benchmark/bench_per_token_group_quant_8bit.py b/python/sglang/jit_kernel/benchmark/bench_per_token_group_quant_8bit.py
index a5a3c392b..549ab0fa5 100644
--- a/python/sglang/jit_kernel/benchmark/bench_per_token_group_quant_8bit.py
+++ b/python/sglang/jit_kernel/benchmark/bench_per_token_group_quant_8bit.py
@@ -20,7 +20,7 @@ from sglang.srt.utils.bench_utils import bench_kineto
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.utils import is_in_ci
 
-register_cuda_ci(est_time=13, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=13, suite="base-b-kernel-benchmark-1-gpu-large")
 
 IS_CI = is_in_ci()
 
diff --git a/python/sglang/jit_kernel/benchmark/bench_qknorm.py b/python/sglang/jit_kernel/benchmark/bench_qknorm.py
index e5458385c..e7c052ab1 100644
--- a/python/sglang/jit_kernel/benchmark/bench_qknorm.py
+++ b/python/sglang/jit_kernel/benchmark/bench_qknorm.py
@@ -15,7 +15,7 @@ from sglang.jit_kernel.norm import fused_inplace_qknorm
 from sglang.srt.utils import get_current_device_stream_fast
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=10, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=10, suite="base-b-kernel-benchmark-1-gpu-large")
 
 alt_stream = torch.cuda.Stream()
 
diff --git a/python/sglang/jit_kernel/benchmark/bench_qknorm_across_heads.py b/python/sglang/jit_kernel/benchmark/bench_qknorm_across_heads.py
index 9bd05f7fc..f969db429 100644
--- a/python/sglang/jit_kernel/benchmark/bench_qknorm_across_heads.py
+++ b/python/sglang/jit_kernel/benchmark/bench_qknorm_across_heads.py
@@ -12,7 +12,7 @@ from sglang.srt.utils import get_current_device_stream_fast
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.utils import is_in_ci
 
-register_cuda_ci(est_time=12, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=12, suite="base-b-kernel-benchmark-1-gpu-large")
 
 IS_CI = is_in_ci()
 
diff --git a/python/sglang/jit_kernel/benchmark/bench_renorm.py b/python/sglang/jit_kernel/benchmark/bench_renorm.py
index f65a615ac..500c9b1b1 100644
--- a/python/sglang/jit_kernel/benchmark/bench_renorm.py
+++ b/python/sglang/jit_kernel/benchmark/bench_renorm.py
@@ -9,7 +9,7 @@ from sglang.jit_kernel.benchmark.utils import run_benchmark_no_cudagraph
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.utils import is_in_ci
 
-register_cuda_ci(est_time=5, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=5, suite="base-b-kernel-benchmark-1-gpu-large")
 
 
 def torch_top_k_renorm_probs(probs, top_k):
diff --git a/python/sglang/jit_kernel/benchmark/bench_resolve_future_token_ids.py b/python/sglang/jit_kernel/benchmark/bench_resolve_future_token_ids.py
index 5551d2968..b0334c6d3 100644
--- a/python/sglang/jit_kernel/benchmark/bench_resolve_future_token_ids.py
+++ b/python/sglang/jit_kernel/benchmark/bench_resolve_future_token_ids.py
@@ -13,7 +13,7 @@ from sglang.jit_kernel.resolve_future_token_ids import resolve_future_token_ids_
 from sglang.srt.utils import get_compiler_backend
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=10, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=10, suite="base-b-kernel-benchmark-1-gpu-large")
 register_amd_ci(est_time=10, suite="jit-kernel-unit-test-amd")
 
 SIZE_LIST = get_benchmark_range(
diff --git a/python/sglang/jit_kernel/benchmark/bench_rope.py b/python/sglang/jit_kernel/benchmark/bench_rope.py
index afe591185..4d578c239 100644
--- a/python/sglang/jit_kernel/benchmark/bench_rope.py
+++ b/python/sglang/jit_kernel/benchmark/bench_rope.py
@@ -12,7 +12,7 @@ from sglang.jit_kernel.benchmark.utils import (
 )
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=6, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=6, suite="base-b-kernel-benchmark-1-gpu-large")
 
 MAX_SEQ_LEN = 131072
 ROPE_BASE = 10000.0
diff --git a/python/sglang/jit_kernel/benchmark/bench_set_mla_kv_buffer.py b/python/sglang/jit_kernel/benchmark/bench_set_mla_kv_buffer.py
index 1ec1e5d88..58a0d9d9a 100644
--- a/python/sglang/jit_kernel/benchmark/bench_set_mla_kv_buffer.py
+++ b/python/sglang/jit_kernel/benchmark/bench_set_mla_kv_buffer.py
@@ -26,7 +26,7 @@ from sglang.srt.mem_cache.utils import set_mla_kv_buffer_kernel as sglang_triton
 from sglang.srt.mem_cache.utils import set_mla_kv_buffer_triton as sglang_wrapper
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=9, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=9, suite="base-b-kernel-benchmark-1-gpu-large")
 
 
 def _triton_baseline(kv_buffer, loc, cache_k_nope, cache_k_rope):
diff --git a/python/sglang/jit_kernel/benchmark/bench_store_cache.py b/python/sglang/jit_kernel/benchmark/bench_store_cache.py
index f1399ff0e..7ce0544ad 100644
--- a/python/sglang/jit_kernel/benchmark/bench_store_cache.py
+++ b/python/sglang/jit_kernel/benchmark/bench_store_cache.py
@@ -14,7 +14,7 @@ from sglang.jit_kernel.benchmark.utils import (
 from sglang.jit_kernel.kvcache import store_cache
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=9, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=9, suite="base-b-kernel-benchmark-1-gpu-large")
 
 
 def sglang_jit_store_cache(
diff --git a/python/sglang/jit_kernel/benchmark/bench_tp_qknorm.py b/python/sglang/jit_kernel/benchmark/bench_tp_qknorm.py
index d63545b10..b3f4d44cd 100644
--- a/python/sglang/jit_kernel/benchmark/bench_tp_qknorm.py
+++ b/python/sglang/jit_kernel/benchmark/bench_tp_qknorm.py
@@ -19,7 +19,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(
     est_time=120,
-    suite="stage-b-kernel-benchmark-1-gpu-large",
+    suite="base-b-kernel-benchmark-1-gpu-large",
     disabled="requires multi-GPU, self-skips in CI",
 )
 
diff --git a/python/sglang/jit_kernel/benchmark/diffusion/bench_diffusion_nvfp4_scaled_mm.py b/python/sglang/jit_kernel/benchmark/diffusion/bench_diffusion_nvfp4_scaled_mm.py
index b6eaaf1f6..35c727dec 100644
--- a/python/sglang/jit_kernel/benchmark/diffusion/bench_diffusion_nvfp4_scaled_mm.py
+++ b/python/sglang/jit_kernel/benchmark/diffusion/bench_diffusion_nvfp4_scaled_mm.py
@@ -17,7 +17,7 @@ from sglang.utils import is_in_ci
 
 register_cuda_ci(
     est_time=120,
-    suite="stage-b-kernel-benchmark-1-gpu-large",
+    suite="base-b-kernel-benchmark-1-gpu-large",
     disabled="standalone diffusion NVFP4 benchmark",
 )
 
diff --git a/python/sglang/jit_kernel/benchmark/diffusion/bench_fused_norm_scale_shift.py b/python/sglang/jit_kernel/benchmark/diffusion/bench_fused_norm_scale_shift.py
index 759241a6d..07bf7926d 100644
--- a/python/sglang/jit_kernel/benchmark/diffusion/bench_fused_norm_scale_shift.py
+++ b/python/sglang/jit_kernel/benchmark/diffusion/bench_fused_norm_scale_shift.py
@@ -20,7 +20,7 @@ from sglang.utils import is_in_ci
 
 register_cuda_ci(
     est_time=17,
-    suite="stage-b-kernel-benchmark-1-gpu-large",
+    suite="base-b-kernel-benchmark-1-gpu-large",
     disabled="Temporarily skipped to unblock flashinfer upgrade. Ref: https://github.com/sgl-project/sglang/actions/runs/23735552939/job/69139238979?pr=21422",
 )
 
diff --git a/python/sglang/jit_kernel/benchmark/diffusion/bench_group_norm_silu.py b/python/sglang/jit_kernel/benchmark/diffusion/bench_group_norm_silu.py
index 0bd48a307..4a6e0f4c6 100644
--- a/python/sglang/jit_kernel/benchmark/diffusion/bench_group_norm_silu.py
+++ b/python/sglang/jit_kernel/benchmark/diffusion/bench_group_norm_silu.py
@@ -16,7 +16,7 @@ from sglang.utils import is_in_ci
 
 register_cuda_ci(
     est_time=45,
-    suite="stage-b-kernel-benchmark-1-gpu-large",
+    suite="base-b-kernel-benchmark-1-gpu-large",
     disabled="standalone benchmark",
 )
 
diff --git a/python/sglang/jit_kernel/benchmark/diffusion/bench_norm_impls.py b/python/sglang/jit_kernel/benchmark/diffusion/bench_norm_impls.py
index 557157f7f..7f0cb058d 100644
--- a/python/sglang/jit_kernel/benchmark/diffusion/bench_norm_impls.py
+++ b/python/sglang/jit_kernel/benchmark/diffusion/bench_norm_impls.py
@@ -24,7 +24,7 @@ from sglang.utils import is_in_ci
 
 register_cuda_ci(
     est_time=120,
-    suite="stage-b-kernel-benchmark-1-gpu-large",
+    suite="base-b-kernel-benchmark-1-gpu-large",
     disabled="self-skips in CI, standalone tool",
 )
 
diff --git a/python/sglang/jit_kernel/benchmark/diffusion/bench_qknorm_rope.py b/python/sglang/jit_kernel/benchmark/diffusion/bench_qknorm_rope.py
index 8d60097bb..46cea6e8e 100644
--- a/python/sglang/jit_kernel/benchmark/diffusion/bench_qknorm_rope.py
+++ b/python/sglang/jit_kernel/benchmark/diffusion/bench_qknorm_rope.py
@@ -13,7 +13,7 @@ from sglang.jit_kernel.benchmark.utils import (
 )
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=13, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=13, suite="base-b-kernel-benchmark-1-gpu-large")
 
 MAX_SEQ_LEN = 131072
 ROPE_BASE = 10000.0
diff --git a/python/sglang/jit_kernel/benchmark/diffusion/bench_qwen_image_modulation.py b/python/sglang/jit_kernel/benchmark/diffusion/bench_qwen_image_modulation.py
index c4713d56a..f87f3e3b3 100644
--- a/python/sglang/jit_kernel/benchmark/diffusion/bench_qwen_image_modulation.py
+++ b/python/sglang/jit_kernel/benchmark/diffusion/bench_qwen_image_modulation.py
@@ -12,7 +12,7 @@ from sglang.jit_kernel.diffusion.triton.scale_shift import (
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.utils import is_in_ci
 
-register_cuda_ci(est_time=13, suite="stage-b-kernel-benchmark-1-gpu-large")
+register_cuda_ci(est_time=13, suite="base-b-kernel-benchmark-1-gpu-large")
 
 if is_in_ci():
     B_RANGE, S_RANGE, D_RANGE = [1], [128], [3072]
diff --git a/python/sglang/jit_kernel/tests/deepseek_v4/test_c128_v2.py b/python/sglang/jit_kernel/tests/deepseek_v4/test_c128_v2.py
index e17de0855..5ec3462c0 100644
--- a/python/sglang/jit_kernel/tests/deepseek_v4/test_c128_v2.py
+++ b/python/sglang/jit_kernel/tests/deepseek_v4/test_c128_v2.py
@@ -18,7 +18,7 @@ from sglang.jit_kernel.tests.deepseek_v4.common import (
     to_seq_extend,
 )
 
-register_cuda_ci(est_time=30, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=30, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=30, suite="nightly-kernel-1-gpu", nightly=True)
 
 Context = Union[LegacyContext, PagedContext]
diff --git a/python/sglang/jit_kernel/tests/deepseek_v4/test_c4_v2.py b/python/sglang/jit_kernel/tests/deepseek_v4/test_c4_v2.py
index 23a513da1..7f8e7c06d 100644
--- a/python/sglang/jit_kernel/tests/deepseek_v4/test_c4_v2.py
+++ b/python/sglang/jit_kernel/tests/deepseek_v4/test_c4_v2.py
@@ -18,7 +18,7 @@ from sglang.jit_kernel.tests.deepseek_v4.common import (
     to_seq_extend,
 )
 
-register_cuda_ci(est_time=30, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=30, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=30, suite="nightly-kernel-1-gpu", nightly=True)
 
 Context = Union[LegacyContext, PagedContext]
diff --git a/python/sglang/jit_kernel/tests/diffusion/test_diffusion_modelopt_fp8_scaled_mm.py b/python/sglang/jit_kernel/tests/diffusion/test_diffusion_modelopt_fp8_scaled_mm.py
index 573f80f8a..8ca94e371 100644
--- a/python/sglang/jit_kernel/tests/diffusion/test_diffusion_modelopt_fp8_scaled_mm.py
+++ b/python/sglang/jit_kernel/tests/diffusion/test_diffusion_modelopt_fp8_scaled_mm.py
@@ -14,7 +14,7 @@ from sglang.srt.layers.quantization.fp8_utils import (
 )
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=20, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=20, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=80, suite="nightly-kernel-1-gpu", nightly=True)
 
 DEVICE = "cuda"
diff --git a/python/sglang/jit_kernel/tests/diffusion/test_diffusion_nvfp4_scaled_mm.py b/python/sglang/jit_kernel/tests/diffusion/test_diffusion_nvfp4_scaled_mm.py
index e2ead525a..2122601cb 100644
--- a/python/sglang/jit_kernel/tests/diffusion/test_diffusion_nvfp4_scaled_mm.py
+++ b/python/sglang/jit_kernel/tests/diffusion/test_diffusion_nvfp4_scaled_mm.py
@@ -17,7 +17,7 @@ from sglang.srt.layers.quantization.modelopt_quant import pad_nvfp4_weight
 from sglang.test.ci.ci_register import register_cuda_ci
 
 # B200-only correctness coverage for diffusion NVFP4 scaled mm.
-register_cuda_ci(est_time=15, suite="stage-b-kernel-unit-1-gpu-b200")
+register_cuda_ci(est_time=15, suite="base-b-kernel-unit-1-gpu-b200")
 
 DEVICE = "cuda"
 DTYPE = torch.bfloat16
diff --git a/python/sglang/jit_kernel/tests/diffusion/test_fused_norm_scale_shift.py b/python/sglang/jit_kernel/tests/diffusion/test_fused_norm_scale_shift.py
index 87cf70a5b..9a0f87c23 100644
--- a/python/sglang/jit_kernel/tests/diffusion/test_fused_norm_scale_shift.py
+++ b/python/sglang/jit_kernel/tests/diffusion/test_fused_norm_scale_shift.py
@@ -13,7 +13,7 @@ from sglang.jit_kernel.diffusion.cutedsl.scale_residual_norm_scale_shift import
 )
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=28, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=28, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 DEVICE = "cuda"
diff --git a/python/sglang/jit_kernel/tests/diffusion/test_group_norm_silu.py b/python/sglang/jit_kernel/tests/diffusion/test_group_norm_silu.py
index c2bb57d0c..63fde24bf 100644
--- a/python/sglang/jit_kernel/tests/diffusion/test_group_norm_silu.py
+++ b/python/sglang/jit_kernel/tests/diffusion/test_group_norm_silu.py
@@ -9,7 +9,7 @@ from sglang.jit_kernel.diffusion.group_norm_silu import apply_group_norm_silu
 from sglang.jit_kernel.diffusion.triton.group_norm_silu import triton_group_norm_silu
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=8, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=8, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 DEVICE = "cuda"
diff --git a/python/sglang/jit_kernel/tests/diffusion/test_qknorm_rope.py b/python/sglang/jit_kernel/tests/diffusion/test_qknorm_rope.py
index f12a6ae8e..0f3370ffc 100644
--- a/python/sglang/jit_kernel/tests/diffusion/test_qknorm_rope.py
+++ b/python/sglang/jit_kernel/tests/diffusion/test_qknorm_rope.py
@@ -8,7 +8,7 @@ import triton
 from sglang.jit_kernel.utils import get_ci_test_range
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=44, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=44, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=176, suite="nightly-kernel-1-gpu", nightly=True)
 
 DEVICE = "cuda"
diff --git a/python/sglang/jit_kernel/tests/diffusion/test_qwen_image_modulation.py b/python/sglang/jit_kernel/tests/diffusion/test_qwen_image_modulation.py
index dce8ce947..0cede04c9 100644
--- a/python/sglang/jit_kernel/tests/diffusion/test_qwen_image_modulation.py
+++ b/python/sglang/jit_kernel/tests/diffusion/test_qwen_image_modulation.py
@@ -12,7 +12,7 @@ from sglang.jit_kernel.diffusion.triton.scale_shift import (
 from sglang.jit_kernel.utils import get_ci_test_range
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=15, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=15, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 DEVICE = "cuda"
diff --git a/python/sglang/jit_kernel/tests/test_activation.py b/python/sglang/jit_kernel/tests/test_activation.py
index 4ad39b269..e384f31f3 100644
--- a/python/sglang/jit_kernel/tests/test_activation.py
+++ b/python/sglang/jit_kernel/tests/test_activation.py
@@ -8,7 +8,7 @@ from sglang.jit_kernel.activation import SUPPORTED_ACTIVATIONS, run_activation
 from sglang.jit_kernel.utils import get_ci_test_range
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=20, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=20, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=30, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_add_constant.py b/python/sglang/jit_kernel/tests/test_add_constant.py
index cad9ac3ab..77cb31fb9 100644
--- a/python/sglang/jit_kernel/tests/test_add_constant.py
+++ b/python/sglang/jit_kernel/tests/test_add_constant.py
@@ -6,7 +6,7 @@ import torch
 from sglang.jit_kernel.add_constant import add_constant
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=45, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=45, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=180, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_awq_dequantize.py b/python/sglang/jit_kernel/tests/test_awq_dequantize.py
index debd97729..19d22e669 100644
--- a/python/sglang/jit_kernel/tests/test_awq_dequantize.py
+++ b/python/sglang/jit_kernel/tests/test_awq_dequantize.py
@@ -7,7 +7,7 @@ import torch
 from sglang.jit_kernel.awq_dequantize import awq_dequantize as jit_awq_dequantize
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=9, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=9, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 try:
diff --git a/python/sglang/jit_kernel/tests/test_awq_marlin_moe_repack.py b/python/sglang/jit_kernel/tests/test_awq_marlin_moe_repack.py
index 2a0f9354e..7148b057c 100644
--- a/python/sglang/jit_kernel/tests/test_awq_marlin_moe_repack.py
+++ b/python/sglang/jit_kernel/tests/test_awq_marlin_moe_repack.py
@@ -11,7 +11,7 @@ from sglang.jit_kernel.awq_marlin_repack import (
 from sglang.srt.layers.quantization.utils import pack_cols, quantize_weights
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=10, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=10, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_awq_marlin_repack.py b/python/sglang/jit_kernel/tests/test_awq_marlin_repack.py
index 35c23922d..4b0746b85 100644
--- a/python/sglang/jit_kernel/tests/test_awq_marlin_repack.py
+++ b/python/sglang/jit_kernel/tests/test_awq_marlin_repack.py
@@ -12,7 +12,7 @@ from sglang.srt.layers.quantization.utils import pack_cols, quantize_weights
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_marlin_utils import get_weight_perm, marlin_weights
 
-register_cuda_ci(est_time=10, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=10, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_clamp_position.py b/python/sglang/jit_kernel/tests/test_clamp_position.py
index cb3ec6ce5..4e8cc1e4f 100644
--- a/python/sglang/jit_kernel/tests/test_clamp_position.py
+++ b/python/sglang/jit_kernel/tests/test_clamp_position.py
@@ -6,7 +6,7 @@ import torch
 from sglang.jit_kernel.clamp_position import clamp_position_cuda
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=12, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=12, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_concat_mla.py b/python/sglang/jit_kernel/tests/test_concat_mla.py
index 9ecfb654c..ceb70d012 100644
--- a/python/sglang/jit_kernel/tests/test_concat_mla.py
+++ b/python/sglang/jit_kernel/tests/test_concat_mla.py
@@ -7,7 +7,7 @@ import triton
 
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=17, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=17, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_custom_all_reduce.py b/python/sglang/jit_kernel/tests/test_custom_all_reduce.py
index 96a61bfbf..a36c05209 100644
--- a/python/sglang/jit_kernel/tests/test_custom_all_reduce.py
+++ b/python/sglang/jit_kernel/tests/test_custom_all_reduce.py
@@ -38,7 +38,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(
     est_time=300,
-    suite="stage-b-kernel-unit-8-gpu-h200",
+    suite="base-b-kernel-unit-8-gpu-h200",
 )
 register_cuda_ci(
     est_time=300,
diff --git a/python/sglang/jit_kernel/tests/test_cutedsl_gdn.py b/python/sglang/jit_kernel/tests/test_cutedsl_gdn.py
index 83e47024c..fb6700d5d 100644
--- a/python/sglang/jit_kernel/tests/test_cutedsl_gdn.py
+++ b/python/sglang/jit_kernel/tests/test_cutedsl_gdn.py
@@ -29,7 +29,7 @@ try:
 except ImportError:
     TRITON_AVAILABLE = False
 
-register_cuda_ci(est_time=5, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=5, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_flash_attention_4.py b/python/sglang/jit_kernel/tests/test_flash_attention_4.py
index 81b0f0b23..95bb978af 100644
--- a/python/sglang/jit_kernel/tests/test_flash_attention_4.py
+++ b/python/sglang/jit_kernel/tests/test_flash_attention_4.py
@@ -14,7 +14,7 @@ from einops import rearrange, repeat
 from sglang.jit_kernel.flash_attention import flash_attn_varlen_func
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=120, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=120, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=900, suite="nightly-kernel-1-gpu", nightly=True)
 
 # Skip this test on Hopper machine
diff --git a/python/sglang/jit_kernel/tests/test_fused_add_rmsnorm.py b/python/sglang/jit_kernel/tests/test_fused_add_rmsnorm.py
index 49636c4f5..2caf7b0dd 100644
--- a/python/sglang/jit_kernel/tests/test_fused_add_rmsnorm.py
+++ b/python/sglang/jit_kernel/tests/test_fused_add_rmsnorm.py
@@ -7,7 +7,7 @@ import torch
 from sglang.jit_kernel.utils import get_ci_test_range
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=5, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=5, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_fused_metadata_copy.py b/python/sglang/jit_kernel/tests/test_fused_metadata_copy.py
index f0fb78c1f..4daeae01c 100644
--- a/python/sglang/jit_kernel/tests/test_fused_metadata_copy.py
+++ b/python/sglang/jit_kernel/tests/test_fused_metadata_copy.py
@@ -16,7 +16,7 @@ import torch
 
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=100, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=100, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=400, suite="nightly-kernel-1-gpu", nightly=True)
 
 # =============================================================================
diff --git a/python/sglang/jit_kernel/tests/test_fused_store_index_cache.py b/python/sglang/jit_kernel/tests/test_fused_store_index_cache.py
index 5766f3168..6dc382166 100644
--- a/python/sglang/jit_kernel/tests/test_fused_store_index_cache.py
+++ b/python/sglang/jit_kernel/tests/test_fused_store_index_cache.py
@@ -48,7 +48,7 @@ try:
 except ImportError:
     _is_fp8_fnuz = False
 
-register_cuda_ci(est_time=24, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=24, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 PAGE_SIZE = 64
diff --git a/python/sglang/jit_kernel/tests/test_fused_verify_triton_gdn.py b/python/sglang/jit_kernel/tests/test_fused_verify_triton_gdn.py
index 08db51289..ec9ba72dc 100644
--- a/python/sglang/jit_kernel/tests/test_fused_verify_triton_gdn.py
+++ b/python/sglang/jit_kernel/tests/test_fused_verify_triton_gdn.py
@@ -26,7 +26,7 @@ try:
 except ImportError:
     KERNELS_AVAILABLE = False
 
-register_cuda_ci(est_time=6, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=6, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_gptq_marlin.py b/python/sglang/jit_kernel/tests/test_gptq_marlin.py
index a811ee654..379d1132a 100644
--- a/python/sglang/jit_kernel/tests/test_gptq_marlin.py
+++ b/python/sglang/jit_kernel/tests/test_gptq_marlin.py
@@ -9,7 +9,7 @@ from sglang.srt.layers.quantization.marlin_utils import marlin_make_workspace
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_marlin_utils import awq_marlin_quantize, marlin_quantize
 
-register_cuda_ci(est_time=13, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=13, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 MNK_FACTORS = [
diff --git a/python/sglang/jit_kernel/tests/test_gptq_marlin_repack.py b/python/sglang/jit_kernel/tests/test_gptq_marlin_repack.py
index 4bcbc0bf3..a94062e17 100644
--- a/python/sglang/jit_kernel/tests/test_gptq_marlin_repack.py
+++ b/python/sglang/jit_kernel/tests/test_gptq_marlin_repack.py
@@ -13,7 +13,7 @@ from sglang.srt.layers.quantization.utils import (
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_marlin_utils import get_weight_perm, marlin_weights
 
-register_cuda_ci(est_time=16, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=16, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 MARLIN_K_CHUNKS = [128]
diff --git a/python/sglang/jit_kernel/tests/test_grouped_topk.py b/python/sglang/jit_kernel/tests/test_grouped_topk.py
index 5c3289094..e47d04609 100644
--- a/python/sglang/jit_kernel/tests/test_grouped_topk.py
+++ b/python/sglang/jit_kernel/tests/test_grouped_topk.py
@@ -9,7 +9,7 @@ from sglang.jit_kernel.utils import get_ci_test_range
 from sglang.srt.layers.moe.topk import biased_grouped_topk_impl
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=30, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=30, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_hadamard_jit.py b/python/sglang/jit_kernel/tests/test_hadamard_jit.py
index cc03a01ae..208a6b1ae 100644
--- a/python/sglang/jit_kernel/tests/test_hadamard_jit.py
+++ b/python/sglang/jit_kernel/tests/test_hadamard_jit.py
@@ -16,7 +16,7 @@ from sglang.jit_kernel.hadamard import (
 )
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=128, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=128, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=512, suite="nightly-kernel-1-gpu", nightly=True)
 
 # Exact M×N Hadamard matrices (±1 entries) copied from
diff --git a/python/sglang/jit_kernel/tests/test_hicache.py b/python/sglang/jit_kernel/tests/test_hicache.py
index b6059b2c1..527aeefe7 100644
--- a/python/sglang/jit_kernel/tests/test_hicache.py
+++ b/python/sglang/jit_kernel/tests/test_hicache.py
@@ -13,7 +13,7 @@ from sglang.srt.mem_cache.memory_pool_host import (
 from sglang.srt.utils import is_cuda, is_hip, is_npu, is_xpu
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=10, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=10, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 pytestmark = pytest.mark.skipif(
diff --git a/python/sglang/jit_kernel/tests/test_hisparse.py b/python/sglang/jit_kernel/tests/test_hisparse.py
index 190f22355..c4b729f5e 100644
--- a/python/sglang/jit_kernel/tests/test_hisparse.py
+++ b/python/sglang/jit_kernel/tests/test_hisparse.py
@@ -7,7 +7,7 @@ from sglang.jit_kernel.hisparse import load_cache_to_device_buffer_mla
 from sglang.srt.utils import is_cuda, is_hip, is_npu, is_xpu
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=10, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=10, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 pytestmark = pytest.mark.skipif(
diff --git a/python/sglang/jit_kernel/tests/test_mla_kv_pack_quantize_fp8.py b/python/sglang/jit_kernel/tests/test_mla_kv_pack_quantize_fp8.py
index bfb88067a..b78c42fd2 100644
--- a/python/sglang/jit_kernel/tests/test_mla_kv_pack_quantize_fp8.py
+++ b/python/sglang/jit_kernel/tests/test_mla_kv_pack_quantize_fp8.py
@@ -7,7 +7,7 @@ from sglang.jit_kernel.mla_kv_pack_quantize_fp8 import mla_kv_pack_quantize_fp8
 from sglang.jit_kernel.utils import get_ci_test_range
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=60, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=60, suite="base-b-kernel-unit-1-gpu-large")
 
 DEVICE = "cuda"
 
diff --git a/python/sglang/jit_kernel/tests/test_moe_align_block_size.py b/python/sglang/jit_kernel/tests/test_moe_align_block_size.py
index 92905058a..a84ba8e40 100644
--- a/python/sglang/jit_kernel/tests/test_moe_align_block_size.py
+++ b/python/sglang/jit_kernel/tests/test_moe_align_block_size.py
@@ -9,7 +9,7 @@ import triton.language as tl
 from sglang.jit_kernel.moe_align import moe_align_block_size
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=28, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=28, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_moe_lora_align_block_size.py b/python/sglang/jit_kernel/tests/test_moe_lora_align_block_size.py
index 48688e5da..b2e721cc9 100644
--- a/python/sglang/jit_kernel/tests/test_moe_lora_align_block_size.py
+++ b/python/sglang/jit_kernel/tests/test_moe_lora_align_block_size.py
@@ -11,7 +11,7 @@ import torch
 from sglang.jit_kernel.moe_lora_align import moe_lora_align_block_size
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=28, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=28, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_moe_wna16_marlin.py b/python/sglang/jit_kernel/tests/test_moe_wna16_marlin.py
index 5100aac5c..54852f1a9 100644
--- a/python/sglang/jit_kernel/tests/test_moe_wna16_marlin.py
+++ b/python/sglang/jit_kernel/tests/test_moe_wna16_marlin.py
@@ -10,7 +10,7 @@ from sglang.srt.layers.moe.fused_moe_triton import moe_align_block_size
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_marlin_utils import awq_marlin_quantize, marlin_quantize
 
-register_cuda_ci(est_time=10, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=10, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_mxfp8_moe.py b/python/sglang/jit_kernel/tests/test_mxfp8_moe.py
index 12157b7aa..4ba53351a 100644
--- a/python/sglang/jit_kernel/tests/test_mxfp8_moe.py
+++ b/python/sglang/jit_kernel/tests/test_mxfp8_moe.py
@@ -10,7 +10,7 @@ from sglang.jit_kernel.mxfp8 import (
 )
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=5, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=5, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_nvfp4_blockwise_moe.py b/python/sglang/jit_kernel/tests/test_nvfp4_blockwise_moe.py
index 864636050..d188ef268 100644
--- a/python/sglang/jit_kernel/tests/test_nvfp4_blockwise_moe.py
+++ b/python/sglang/jit_kernel/tests/test_nvfp4_blockwise_moe.py
@@ -10,7 +10,7 @@ from sglang.jit_kernel.nvfp4 import (
 )
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=5, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=5, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 FLOAT4_E2M1_MAX = 6.0
diff --git a/python/sglang/jit_kernel/tests/test_nvfp4_gemm.py b/python/sglang/jit_kernel/tests/test_nvfp4_gemm.py
index 21c14d5b1..4e55fe40b 100644
--- a/python/sglang/jit_kernel/tests/test_nvfp4_gemm.py
+++ b/python/sglang/jit_kernel/tests/test_nvfp4_gemm.py
@@ -6,7 +6,7 @@ import torch
 from sglang.jit_kernel.nvfp4 import cutlass_scaled_fp4_mm, scaled_fp4_quant
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=5, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=5, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_nvfp4_quant.py b/python/sglang/jit_kernel/tests/test_nvfp4_quant.py
index 5b7dfbd0f..293d93ead 100644
--- a/python/sglang/jit_kernel/tests/test_nvfp4_quant.py
+++ b/python/sglang/jit_kernel/tests/test_nvfp4_quant.py
@@ -16,7 +16,7 @@ except Exception:
 
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=5, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=5, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_per_tensor_quant_fp8.py b/python/sglang/jit_kernel/tests/test_per_tensor_quant_fp8.py
index cac76d03a..2e3acea3a 100644
--- a/python/sglang/jit_kernel/tests/test_per_tensor_quant_fp8.py
+++ b/python/sglang/jit_kernel/tests/test_per_tensor_quant_fp8.py
@@ -8,7 +8,7 @@ import torch
 from sglang.jit_kernel.per_tensor_quant_fp8 import per_tensor_quant_fp8
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=16, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=16, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 try:
diff --git a/python/sglang/jit_kernel/tests/test_per_token_group_quant_8bit.py b/python/sglang/jit_kernel/tests/test_per_token_group_quant_8bit.py
index 8b0452d7d..f370cb19a 100644
--- a/python/sglang/jit_kernel/tests/test_per_token_group_quant_8bit.py
+++ b/python/sglang/jit_kernel/tests/test_per_token_group_quant_8bit.py
@@ -25,7 +25,7 @@ from sglang.srt.layers.quantization.fp8_kernel import (
 )
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=16, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=16, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 configs = list(
diff --git a/python/sglang/jit_kernel/tests/test_pos_enc.py b/python/sglang/jit_kernel/tests/test_pos_enc.py
index 8b4fcd146..cdeb3f85f 100644
--- a/python/sglang/jit_kernel/tests/test_pos_enc.py
+++ b/python/sglang/jit_kernel/tests/test_pos_enc.py
@@ -10,7 +10,7 @@ import triton.language as tl
 from sglang.jit_kernel.rope import rotary_embedding
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=18, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=18, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_qknorm.py b/python/sglang/jit_kernel/tests/test_qknorm.py
index 4b85b5c52..b0026997c 100644
--- a/python/sglang/jit_kernel/tests/test_qknorm.py
+++ b/python/sglang/jit_kernel/tests/test_qknorm.py
@@ -8,7 +8,7 @@ import triton
 from sglang.jit_kernel.utils import get_ci_test_range
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=37, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=37, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=148, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_qknorm_across_heads.py b/python/sglang/jit_kernel/tests/test_qknorm_across_heads.py
index 81c8a1ead..a62887a73 100644
--- a/python/sglang/jit_kernel/tests/test_qknorm_across_heads.py
+++ b/python/sglang/jit_kernel/tests/test_qknorm_across_heads.py
@@ -8,7 +8,7 @@ import triton
 from sglang.jit_kernel.utils import get_ci_test_range
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=15, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=15, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_renorm.py b/python/sglang/jit_kernel/tests/test_renorm.py
index d3ef6ce19..119fcd80d 100644
--- a/python/sglang/jit_kernel/tests/test_renorm.py
+++ b/python/sglang/jit_kernel/tests/test_renorm.py
@@ -9,7 +9,7 @@ import torch
 
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=6, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=6, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_resolve_future_token_ids.py b/python/sglang/jit_kernel/tests/test_resolve_future_token_ids.py
index cfb8597f6..062625ce2 100644
--- a/python/sglang/jit_kernel/tests/test_resolve_future_token_ids.py
+++ b/python/sglang/jit_kernel/tests/test_resolve_future_token_ids.py
@@ -6,7 +6,7 @@ import torch
 from sglang.jit_kernel.resolve_future_token_ids import resolve_future_token_ids_cuda
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=9, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=9, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_rmsnorm.py b/python/sglang/jit_kernel/tests/test_rmsnorm.py
index 59ce90f29..38b2efc90 100644
--- a/python/sglang/jit_kernel/tests/test_rmsnorm.py
+++ b/python/sglang/jit_kernel/tests/test_rmsnorm.py
@@ -7,7 +7,7 @@ import torch
 from sglang.jit_kernel.utils import get_ci_test_range
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=45, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=45, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=240, suite="nightly-kernel-1-gpu", nightly=True)
 
 
diff --git a/python/sglang/jit_kernel/tests/test_rmsnorm_hf.py b/python/sglang/jit_kernel/tests/test_rmsnorm_hf.py
index 5887bff6c..02162172f 100644
--- a/python/sglang/jit_kernel/tests/test_rmsnorm_hf.py
+++ b/python/sglang/jit_kernel/tests/test_rmsnorm_hf.py
@@ -13,7 +13,7 @@ from sglang.jit_kernel.rmsnorm_hf import (
 from sglang.jit_kernel.utils import get_ci_test_range
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=30, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=30, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 EPS = 1e-5
diff --git a/python/sglang/jit_kernel/tests/test_rope.py b/python/sglang/jit_kernel/tests/test_rope.py
index 62a601f65..28f4e35f5 100644
--- a/python/sglang/jit_kernel/tests/test_rope.py
+++ b/python/sglang/jit_kernel/tests/test_rope.py
@@ -7,7 +7,7 @@ import triton
 from sglang.jit_kernel.utils import get_ci_test_range
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=64, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=64, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=256, suite="nightly-kernel-1-gpu", nightly=True)
 
 DEVICE = "cuda"
diff --git a/python/sglang/jit_kernel/tests/test_set_mla_kv_buffer.py b/python/sglang/jit_kernel/tests/test_set_mla_kv_buffer.py
index f0c395c89..f9aa08724 100644
--- a/python/sglang/jit_kernel/tests/test_set_mla_kv_buffer.py
+++ b/python/sglang/jit_kernel/tests/test_set_mla_kv_buffer.py
@@ -10,7 +10,7 @@ from sglang.jit_kernel.set_mla_kv_buffer import (
 from sglang.jit_kernel.utils import get_ci_test_range
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=30, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=30, suite="base-b-kernel-unit-1-gpu-large")
 
 DEVICE = "cuda"
 CACHE_SIZE = 4096
diff --git a/python/sglang/jit_kernel/tests/test_store_cache.py b/python/sglang/jit_kernel/tests/test_store_cache.py
index a7b9e5be1..16da71362 100644
--- a/python/sglang/jit_kernel/tests/test_store_cache.py
+++ b/python/sglang/jit_kernel/tests/test_store_cache.py
@@ -8,7 +8,7 @@ from sglang.jit_kernel.kvcache import can_use_store_cache, store_cache
 from sglang.jit_kernel.utils import get_ci_test_range
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=28, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=28, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 register_amd_ci(est_time=55, suite="jit-kernel-unit-test-amd")
 
diff --git a/python/sglang/jit_kernel/tests/test_timestep_embedding.py b/python/sglang/jit_kernel/tests/test_timestep_embedding.py
index 1ec5912d4..94db3f4f7 100644
--- a/python/sglang/jit_kernel/tests/test_timestep_embedding.py
+++ b/python/sglang/jit_kernel/tests/test_timestep_embedding.py
@@ -16,7 +16,7 @@ from sglang.jit_kernel.timestep_embedding import (
 from sglang.jit_kernel.utils import get_ci_test_range
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=16, suite="stage-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=16, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
 
 CORRECTNESS_BATCH_SIZES = get_ci_test_range(
diff --git a/python/sglang/jit_kernel/tests/test_tp_qknorm.py b/python/sglang/jit_kernel/tests/test_tp_qknorm.py
index 98a803ca0..854d984d1 100644
--- a/python/sglang/jit_kernel/tests/test_tp_qknorm.py
+++ b/python/sglang/jit_kernel/tests/test_tp_qknorm.py
@@ -16,7 +16,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(
     est_time=300,
-    suite="stage-b-kernel-unit-8-gpu-h200",
+    suite="base-b-kernel-unit-8-gpu-h200",
 )
 register_cuda_ci(
     est_time=300,
diff --git a/python/sglang/test/ci/ci_register.py b/python/sglang/test/ci/ci_register.py
index edf830493..f385f733e 100644
--- a/python/sglang/test/ci/ci_register.py
+++ b/python/sglang/test/ci/ci_register.py
@@ -17,7 +17,7 @@ __all__ = [
 ]
 
 # `suite` stays in positional slot 2 for backward compat with existing
-# `register_cpu_ci(5, "stage-a-test-cpu")` style positional calls. New fields
+# `register_cpu_ci(5, "base-a-test-cpu")` style positional calls. New fields
 # (`stage`, `runner_config`) are kwarg-only.
 _PARAM_ORDER = ("est_time", "suite", "nightly", "disabled")
 _KWARG_ONLY = ("stage", "runner_config")
diff --git a/scripts/ci/utils/compute_partitions.py b/scripts/ci/utils/compute_partitions.py
index 994d877f4..d1f52ce49 100644
--- a/scripts/ci/utils/compute_partitions.py
+++ b/scripts/ci/utils/compute_partitions.py
@@ -2,7 +2,7 @@
 keyed by suite name. Consumed by pr-test.yml stage jobs as
 `fromJson(needs.check-changes.outputs.partitions)['<suite>']`.
 
-    partitions={"stage-b-test-1-gpu-small": {"size": 8, "arr": [0,...,7], "max_parallel": 2}, ...}
+    partitions={"base-b-test-1-gpu-small": {"size": 8, "arr": [0,...,7], "max_parallel": 2}, ...}
 """
 
 import argparse
@@ -34,11 +34,11 @@ HWBackend = _ci_register.HWBackend
 # pr-test-amd.yml / pr-test-npu.yml have their own dispatch.
 _TARGET_BACKENDS = {HWBackend.CUDA, HWBackend.CPU}
 
-# stage-a is the critical-path entry gate; pin its fanout to smoke-coverage
+# base-a is the critical-path entry gate; pin its fanout to smoke-coverage
 # defaults instead of est_time. max_parallel = size (no throttle).
-_STAGE_A_OVERRIDES = {
-    "stage-a-test-cpu": 4,
-    "stage-a-test-1-gpu-small": 1,
+_BASE_A_OVERRIDES = {
+    "base-a-test-cpu": 4,
+    "base-a-test-1-gpu-small": 1,
 }
 
 _REUSABLE_STAGE_USES = "./.github/workflows/_pr-test-stage.yml"
@@ -47,7 +47,7 @@ _REUSABLE_STAGE_USES = "./.github/workflows/_pr-test-stage.yml"
 def load_run_timeouts(pr_test_yml_path: str) -> dict:
     """Map `self_name -> run_timeout_minutes` from one pr-test*.yml. The input
     is required in `_pr-test-stage.yml` -- KeyError surfaces missing.
-    Inline stage-a-test-cpu is skipped (uses `_STAGE_A_OVERRIDES`)."""
+    Inline base-a-test-cpu is skipped (uses `_BASE_A_OVERRIDES`)."""
     with open(pr_test_yml_path) as f:
         wf = yaml.safe_load(f)
     timeouts = {}
@@ -122,7 +122,7 @@ def compute_partitions(
     """
     # Allowlist: stages pr-test.yml dispatches. Stress / weekly /
     # nightly-* live in test/registered/ but pr-test doesn't run them.
-    dispatched_suites = set(run_timeouts) | set(_STAGE_A_OVERRIDES)
+    dispatched_suites = set(run_timeouts) | set(_BASE_A_OVERRIDES)
     suite_tests = defaultdict(list)
     for t in tests:
         if t.backend not in _TARGET_BACKENDS:
@@ -149,8 +149,8 @@ def compute_partitions(
         bias = fit.get("bias", 0.0)
 
         # Each shard pays `bias` once, so size >= coeff*total / (target-bias).
-        if suite in _STAGE_A_OVERRIDES:
-            size = _STAGE_A_OVERRIDES[suite]
+        if suite in _BASE_A_OVERRIDES:
+            size = _BASE_A_OVERRIDES[suite]
             max_parallel = size
         else:
             target = per_shard_target_seconds(suite, run_timeouts)
diff --git a/test/README.md b/test/README.md
index bc7f654b3..414e02b40 100644
--- a/test/README.md
+++ b/test/README.md
@@ -48,14 +48,14 @@ python3 test/registered/core/test_srt_endpoint.py TestSRTEndpoint.test_simple_de
 python3 python/sglang/jit_kernel/tests/test_add_constant.py
 
 # Run a suite
-python3 test/run_suite.py --hw cpu --suite stage-a-test-cpu
-python3 test/run_suite.py --hw cuda --suite stage-a-test-1-gpu-small
+python3 test/run_suite.py --hw cpu --suite base-a-test-cpu
+python3 test/run_suite.py --hw cuda --suite base-a-test-1-gpu-small
 
 # Nightly tests
 python3 test/run_suite.py --hw cuda --suite nightly-1-gpu --nightly
 
 # With auto-partitioning (for parallel CI jobs)
-python3 test/run_suite.py --hw cuda --suite stage-b-test-1-gpu-small \
+python3 test/run_suite.py --hw cuda --suite base-b-test-1-gpu-small \
     --auto-partition-id 0 --auto-partition-size 4
 ```
 
@@ -66,7 +66,7 @@ Every CI-discovered test file must call a registration function at module level:
 ```python
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=80, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=80, stage="base-b", runner_config="1-gpu-small")
 ```
 
 Parameters: `est_time` (seconds), `stage` + `runner_config` (target stage and runner pool from `scripts/ci/runner_configs.yml`), `nightly=True` (nightly-only), `disabled="reason"` (temporarily disable).
@@ -74,8 +74,8 @@ Parameters: `est_time` (seconds), `stage` + `runner_config` (target stage and ru
 Keep `est_time`, `stage`, `runner_config` as **literal values** — `run_suite.py` collects them by AST parsing.
 
 JIT kernel files live outside `test/registered/` but still use registration:
-- Correctness tests: `python/sglang/jit_kernel/tests/test_*.py` → `stage-b-kernel-unit-1-gpu-large`
-- Benchmarks: `python/sglang/jit_kernel/benchmark/bench_*.py` → `stage-b-kernel-benchmark-1-gpu-large`
+- Correctness tests: `python/sglang/jit_kernel/tests/test_*.py` → `base-b-kernel-unit-1-gpu-large`
+- Benchmarks: `python/sglang/jit_kernel/benchmark/bench_*.py` → `base-b-kernel-benchmark-1-gpu-large`
 
 ## Choosing a Suite
 
@@ -83,12 +83,12 @@ Use the lightest suite that meets your test's needs. Full suite tables are in th
 
 | Need | Suite |
 |------|-------|
-| No GPU required | `stage-a-test-cpu` |
-| Small GPU (fits 5090, 32GB) | `stage-b-test-1-gpu-small` (most tests go here) |
-| Large GPU memory or Hopper features | `stage-b-test-1-gpu-large` |
-| JIT kernel correctness | `stage-b-kernel-unit-1-gpu-large` |
-| JIT kernel benchmarks | `stage-b-kernel-benchmark-1-gpu-large` |
-| Multi-GPU (2/4/8) | `stage-b-test-2-gpu-large`, `stage-c-test-*` |
+| No GPU required | `base-a-test-cpu` |
+| Small GPU (fits 5090, 32GB) | `base-b-test-1-gpu-small` (most tests go here) |
+| Large GPU memory or Hopper features | `base-b-test-1-gpu-large` |
+| JIT kernel correctness | `base-b-kernel-unit-1-gpu-large` |
+| JIT kernel benchmarks | `base-b-kernel-benchmark-1-gpu-large` |
+| Multi-GPU (2/4/8) | `base-b-test-2-gpu-large`, `base-c-test-*` |
 | Long-running or experimental | `nightly-*` suites |
 
 ## Steps for Adding a Test
diff --git a/test/registered/4-gpu-models/test_deepseek_v3_cutedsl_4gpu.py b/test/registered/4-gpu-models/test_deepseek_v3_cutedsl_4gpu.py
index 1effb77f3..c59389f40 100644
--- a/test/registered/4-gpu-models/test_deepseek_v3_cutedsl_4gpu.py
+++ b/test/registered/4-gpu-models/test_deepseek_v3_cutedsl_4gpu.py
@@ -14,7 +14,7 @@ from sglang.test.test_utils import (
     try_cached_model,
 )
 
-register_cuda_ci(est_time=1800, stage="stage-c", runner_config="4-gpu-gb200")
+register_cuda_ci(est_time=1800, stage="base-c", runner_config="4-gpu-gb200")
 
 
 class TestDeepseekR1Nvfp4CuteDSLDeepEP(CustomTestCase):
diff --git a/test/registered/4-gpu-models/test_gpt_oss_4gpu.py b/test/registered/4-gpu-models/test_gpt_oss_4gpu.py
index 06a53d646..c792c4d56 100644
--- a/test/registered/4-gpu-models/test_gpt_oss_4gpu.py
+++ b/test/registered/4-gpu-models/test_gpt_oss_4gpu.py
@@ -3,8 +3,8 @@ import unittest
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.gpt_oss_common import BaseTestGptOss
 
-register_cuda_ci(est_time=392, stage="stage-c", runner_config="4-gpu-h100")
-register_cuda_ci(est_time=350, stage="stage-c", runner_config="4-gpu-b200")
+register_cuda_ci(est_time=392, stage="base-c", runner_config="4-gpu-h100")
+register_cuda_ci(est_time=350, stage="base-c", runner_config="4-gpu-b200")
 
 
 class TestGptOss4Gpu(BaseTestGptOss):
diff --git a/test/registered/4-gpu-models/test_qwen35_fp4_mtp_v2.py b/test/registered/4-gpu-models/test_qwen35_fp4_mtp_v2.py
index 6e4c39c42..6e9c8ca6a 100644
--- a/test/registered/4-gpu-models/test_qwen35_fp4_mtp_v2.py
+++ b/test/registered/4-gpu-models/test_qwen35_fp4_mtp_v2.py
@@ -15,7 +15,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=340, stage="stage-c", runner_config="4-gpu-b200")
+register_cuda_ci(est_time=340, stage="base-c", runner_config="4-gpu-b200")
 
 QWEN35_FP4_MODEL = "nvidia/Qwen3.5-397B-A17B-NVFP4"
 ACC_THRESHOLDS = {QWEN35_FP4_MODEL: {"gsm8k": 0.95}}
diff --git a/test/registered/4-gpu-models/test_qwen35_models.py b/test/registered/4-gpu-models/test_qwen35_models.py
index beee76658..434f690c5 100644
--- a/test/registered/4-gpu-models/test_qwen35_models.py
+++ b/test/registered/4-gpu-models/test_qwen35_models.py
@@ -17,7 +17,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=370, stage="stage-c", runner_config="4-gpu-b200")
+register_cuda_ci(est_time=370, stage="base-c", runner_config="4-gpu-b200")
 
 QWEN35_FP4_MODEL = "nvidia/Qwen3.5-397B-A17B-NVFP4"
 ACC_THRESHOLDS = {QWEN35_FP4_MODEL: {"gsm8k": 0.95}}
diff --git a/test/registered/4-gpu-models/test_qwen3_30b.py b/test/registered/4-gpu-models/test_qwen3_30b.py
index 238564599..b81534268 100644
--- a/test/registered/4-gpu-models/test_qwen3_30b.py
+++ b/test/registered/4-gpu-models/test_qwen3_30b.py
@@ -11,7 +11,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=261, stage="stage-c", runner_config="4-gpu-h100")
+register_cuda_ci(est_time=261, stage="base-c", runner_config="4-gpu-h100")
 
 QWEN3_30B_MODEL_PATH = "Qwen/Qwen3-30B-A3B-FP8"
 
diff --git a/test/registered/4-gpu-models/test_qwen3_next_models_mtp.py b/test/registered/4-gpu-models/test_qwen3_next_models_mtp.py
index 1955dedc6..6c8dc6110 100644
--- a/test/registered/4-gpu-models/test_qwen3_next_models_mtp.py
+++ b/test/registered/4-gpu-models/test_qwen3_next_models_mtp.py
@@ -6,7 +6,7 @@ from sglang.test.kits.kl_divergence_kit import KLDivergenceMixin
 from sglang.test.kits.prefix_cache_branching_kit import PrefixCacheBranchingMixin
 from sglang.test.server_fixtures.default_fixture import DefaultServerBase
 
-register_cuda_ci(est_time=290, stage="stage-c", runner_config="4-gpu-h100")
+register_cuda_ci(est_time=290, stage="base-c", runner_config="4-gpu-h100")
 
 QWEN3_NEXT_MODEL = "Qwen/Qwen3-Next-80B-A3B-Instruct"
 
diff --git a/test/registered/8-gpu-models/test_deepseek_v3_mtp.py b/test/registered/8-gpu-models/test_deepseek_v3_mtp.py
index 371184789..9bcb6ebe1 100644
--- a/test/registered/8-gpu-models/test_deepseek_v3_mtp.py
+++ b/test/registered/8-gpu-models/test_deepseek_v3_mtp.py
@@ -17,7 +17,7 @@ from sglang.test.test_utils import (
     write_github_step_summary,
 )
 
-register_cuda_ci(est_time=300, stage="stage-c", runner_config="8-gpu-h200")
+register_cuda_ci(est_time=300, stage="base-c", runner_config="8-gpu-h200")
 
 FULL_DEEPSEEK_V3_MODEL_PATH = "deepseek-ai/DeepSeek-V3-0324"
 
diff --git a/test/registered/8-gpu-models/test_dsa_models_hisparse.py b/test/registered/8-gpu-models/test_dsa_models_hisparse.py
index 73287134f..8bc6575ee 100644
--- a/test/registered/8-gpu-models/test_dsa_models_hisparse.py
+++ b/test/registered/8-gpu-models/test_dsa_models_hisparse.py
@@ -12,9 +12,7 @@ from sglang.test.test_utils import (
     write_github_step_summary,
 )
 
-register_cuda_ci(
-    est_time=720, stage="stage-c", runner_config="8-gpu-h200", nightly=True
-)
+register_cuda_ci(est_time=720, stage="base-c", runner_config="8-gpu-h200", nightly=True)
 
 GLM5_MODEL_PATH = "zai-org/GLM-5-FP8"
 
diff --git a/test/registered/8-gpu-models/test_mimo_models.py b/test/registered/8-gpu-models/test_mimo_models.py
index 41ed0ed44..6d4c65230 100644
--- a/test/registered/8-gpu-models/test_mimo_models.py
+++ b/test/registered/8-gpu-models/test_mimo_models.py
@@ -6,7 +6,7 @@ from sglang.test.kits.spec_decoding_kit import SpecDecodingMixin
 from sglang.test.server_fixtures.default_fixture import DefaultServerBase
 from sglang.test.server_fixtures.mmmu_fixture import MMMUServerBase
 
-register_cuda_ci(est_time=500, stage="stage-c", runner_config="8-gpu-h200")
+register_cuda_ci(est_time=500, stage="base-c", runner_config="8-gpu-h200")
 
 
 class TestMiMoV2Flash(GSM8KMixin, SpecDecodingMixin, DefaultServerBase):
diff --git a/test/registered/8-gpu-models/test_minimax_m25_basic.py b/test/registered/8-gpu-models/test_minimax_m25_basic.py
index b2347e30c..82a3054c0 100644
--- a/test/registered/8-gpu-models/test_minimax_m25_basic.py
+++ b/test/registered/8-gpu-models/test_minimax_m25_basic.py
@@ -14,7 +14,7 @@ from sglang.test.test_utils import (
     write_github_step_summary,
 )
 
-register_cuda_ci(est_time=290, stage="stage-c", runner_config="8-gpu-h200")
+register_cuda_ci(est_time=290, stage="base-c", runner_config="8-gpu-h200")
 
 MINIMAX_M25_MODEL_PATH = "MiniMaxAI/MiniMax-M2.5"
 
diff --git a/test/registered/attention/test_chunk_gated_delta_rule.py b/test/registered/attention/test_chunk_gated_delta_rule.py
index f28a720e8..d3e852963 100644
--- a/test/registered/attention/test_chunk_gated_delta_rule.py
+++ b/test/registered/attention/test_chunk_gated_delta_rule.py
@@ -8,7 +8,7 @@ from sglang.srt.layers.attention.fla.fused_recurrent import (
 )
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=11, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-large")
 
 
 @unittest.skipIf(not torch.cuda.is_available(), "Test requires CUDA")
diff --git a/test/registered/attention/test_create_kvindices.py b/test/registered/attention/test_create_kvindices.py
index c52581c41..898e84e80 100644
--- a/test/registered/attention/test_create_kvindices.py
+++ b/test/registered/attention/test_create_kvindices.py
@@ -9,7 +9,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
 # Triton kernel unit test for KV indices creation
-register_cuda_ci(est_time=7, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=7, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=10, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/attention/test_flash_attention_4.py b/test/registered/attention/test_flash_attention_4.py
index b0210afbc..5ec464099 100644
--- a/test/registered/attention/test_flash_attention_4.py
+++ b/test/registered/attention/test_flash_attention_4.py
@@ -13,7 +13,7 @@ from sglang.test.test_utils import (
 )
 
 # FlashAttention4 integration test (requires SM 100+ / Blackwell B200)
-register_cuda_ci(est_time=260, stage="stage-b", runner_config="4-gpu-b200")
+register_cuda_ci(est_time=260, stage="base-b", runner_config="4-gpu-b200")
 
 
 @unittest.skipIf(get_device_sm() < 100, "Test requires CUDA SM 100 or higher")
diff --git a/test/registered/attention/test_gdn_noncontiguous_stride.py b/test/registered/attention/test_gdn_noncontiguous_stride.py
index 2ba172952..e60a5ef2a 100644
--- a/test/registered/attention/test_gdn_noncontiguous_stride.py
+++ b/test/registered/attention/test_gdn_noncontiguous_stride.py
@@ -14,7 +14,7 @@ from sglang.srt.layers.attention.fla.fused_sigmoid_gating_recurrent import (
 )
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=7, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=7, stage="base-b", runner_config="1-gpu-large")
 
 
 def _make_noncontiguous_ab(batch, num_heads, dtype=torch.bfloat16, device="cuda"):
diff --git a/test/registered/attention/test_kda_kernels.py b/test/registered/attention/test_kda_kernels.py
index b44ef7f01..96cb1e9e8 100644
--- a/test/registered/attention/test_kda_kernels.py
+++ b/test/registered/attention/test_kda_kernels.py
@@ -14,7 +14,7 @@ from sglang.srt.layers.attention.fla.kda import (
 from sglang.srt.utils.common import get_device
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=12, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=12, stage="base-b", runner_config="1-gpu-large")
 
 
 @unittest.skipIf(
diff --git a/test/registered/attention/test_normal_decode_set_metadata.py b/test/registered/attention/test_normal_decode_set_metadata.py
index 4deb2a418..edf5f90fd 100644
--- a/test/registered/attention/test_normal_decode_set_metadata.py
+++ b/test/registered/attention/test_normal_decode_set_metadata.py
@@ -20,8 +20,8 @@ from sglang.srt.mem_cache.swa_memory_pool import SWAKVPool
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-# Register this test for CUDA CI in stage-b (fast attention/kernel tests)
-register_cuda_ci(est_time=11, stage="stage-b", runner_config="1-gpu-large")
+# Register this test for CUDA CI in base-b (fast attention/kernel tests)
+register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-large")
 
 
 def reference_normal_decode_set_metadata(
diff --git a/test/registered/attention/test_triton_attention_backend.py b/test/registered/attention/test_triton_attention_backend.py
index b2aac8dda..17e1f2238 100644
--- a/test/registered/attention/test_triton_attention_backend.py
+++ b/test/registered/attention/test_triton_attention_backend.py
@@ -20,7 +20,7 @@ from sglang.test.test_utils import (
 )
 
 # Triton attention backend integration test with latency benchmark and MMLU eval
-register_cuda_ci(est_time=177, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=177, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=1400, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/attention/test_triton_attention_kernels.py b/test/registered/attention/test_triton_attention_kernels.py
index 8002df396..dfc86cbf3 100644
--- a/test/registered/attention/test_triton_attention_kernels.py
+++ b/test/registered/attention/test_triton_attention_kernels.py
@@ -23,7 +23,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase, is_in_amd_ci
 
 # Triton attention kernel unit tests (decode, extend, prefill)
-register_cuda_ci(est_time=19, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=19, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=30, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/bench_fn/test_bench_serving_reasoning_stream.py b/test/registered/bench_fn/test_bench_serving_reasoning_stream.py
index c33bd2630..d6cf2f591 100644
--- a/test/registered/bench_fn/test_bench_serving_reasoning_stream.py
+++ b/test/registered/bench_fn/test_bench_serving_reasoning_stream.py
@@ -24,7 +24,7 @@ from sglang.bench_serving import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 
 def _free_port() -> int:
diff --git a/test/registered/bench_fn/test_benchmark_datasets_api.py b/test/registered/bench_fn/test_benchmark_datasets_api.py
index 0b9803698..395f9d399 100644
--- a/test/registered/bench_fn/test_benchmark_datasets_api.py
+++ b/test/registered/bench_fn/test_benchmark_datasets_api.py
@@ -26,7 +26,7 @@ from sglang.benchmark.datasets.random import sample_random_requests
 from sglang.benchmark.datasets.sharegpt import sample_sharegpt_requests
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 
 class _DummyTokenTensor:
diff --git a/test/registered/breakable_cuda_graph/test_breakable_cuda_graph.py b/test/registered/breakable_cuda_graph/test_breakable_cuda_graph.py
index aeb391769..9b23bee62 100644
--- a/test/registered/breakable_cuda_graph/test_breakable_cuda_graph.py
+++ b/test/registered/breakable_cuda_graph/test_breakable_cuda_graph.py
@@ -23,7 +23,7 @@ from sglang.test.test_utils import (
 )
 
 # CI Registration — large suite to fit the integration test's server startup.
-register_cuda_ci(est_time=79, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=79, stage="base-b", runner_config="1-gpu-large")
 
 
 def _skip_if_no_cuda(test_func):
diff --git a/test/registered/constrained_decoding/test_constrained_decoding.py b/test/registered/constrained_decoding/test_constrained_decoding.py
index 54c9e01a1..976445b10 100644
--- a/test/registered/constrained_decoding/test_constrained_decoding.py
+++ b/test/registered/constrained_decoding/test_constrained_decoding.py
@@ -13,7 +13,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=120, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=120, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=179, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/core/test_deterministic.py b/test/registered/core/test_deterministic.py
index 97d9c5c8d..b7cdc7814 100644
--- a/test/registered/core/test_deterministic.py
+++ b/test/registered/core/test_deterministic.py
@@ -16,7 +16,7 @@ from sglang.test.test_deterministic_utils import (
 )
 from sglang.test.test_utils import is_in_amd_ci
 
-register_cuda_ci(est_time=207, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=207, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=278, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/core/test_engine_child_pids.py b/test/registered/core/test_engine_child_pids.py
index e9bbf0a92..9cd31b992 100644
--- a/test/registered/core/test_engine_child_pids.py
+++ b/test/registered/core/test_engine_child_pids.py
@@ -20,7 +20,7 @@ from sglang.test.test_utils import (
     CustomTestCase,
 )
 
-register_cuda_ci(est_time=77, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=77, stage="base-b", runner_config="1-gpu-small")
 
 
 class TestEngineChildPids(CustomTestCase):
diff --git a/test/registered/core/test_gemma4_moe_deterministic.py b/test/registered/core/test_gemma4_moe_deterministic.py
index 109af8243..972b79c38 100644
--- a/test/registered/core/test_gemma4_moe_deterministic.py
+++ b/test/registered/core/test_gemma4_moe_deterministic.py
@@ -23,7 +23,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=107, stage="stage-b", runner_config="2-gpu-large")
+register_cuda_ci(est_time=107, stage="base-b", runner_config="2-gpu-large")
 
 
 PROMPT = (
diff --git a/test/registered/core/test_hidden_states.py b/test/registered/core/test_hidden_states.py
index 454393d11..de72a37ce 100644
--- a/test/registered/core/test_hidden_states.py
+++ b/test/registered/core/test_hidden_states.py
@@ -8,7 +8,7 @@ from sglang.srt.utils import get_device, is_hip
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import DEFAULT_SMALL_MODEL_NAME_FOR_TEST, CustomTestCase
 
-register_cuda_ci(est_time=45, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=45, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=55, suite="stage-b-test-1-gpu-small-amd")
 
 _is_hip = is_hip()
diff --git a/test/registered/core/test_mm_process_config.py b/test/registered/core/test_mm_process_config.py
index 6fbdb295d..202cbd4f7 100644
--- a/test/registered/core/test_mm_process_config.py
+++ b/test/registered/core/test_mm_process_config.py
@@ -4,7 +4,7 @@ from unittest.mock import MagicMock, patch
 from sglang.srt.server_args import ServerArgs
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=9, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=1, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/core/test_request_queue_validation.py b/test/registered/core/test_request_queue_validation.py
index 9fb096c7f..3bc9841b9 100644
--- a/test/registered/core/test_request_queue_validation.py
+++ b/test/registered/core/test_request_queue_validation.py
@@ -17,7 +17,7 @@ from sglang.test.test_utils import (
     send_generate_requests,
 )
 
-register_cuda_ci(est_time=53, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=53, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=70, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/core/test_srt_endpoint.py b/test/registered/core/test_srt_endpoint.py
index 1108a005b..749f50d08 100644
--- a/test/registered/core/test_srt_endpoint.py
+++ b/test/registered/core/test_srt_endpoint.py
@@ -28,7 +28,7 @@ from sglang.test.test_utils import (
     run_logprob_check,
 )
 
-register_cuda_ci(est_time=134, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=134, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=130, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/core/test_srt_engine.py b/test/registered/core/test_srt_engine.py
index 86d1ecc10..c849be7b0 100644
--- a/test/registered/core/test_srt_engine.py
+++ b/test/registered/core/test_srt_engine.py
@@ -22,7 +22,7 @@ from sglang.test.test_utils import (
     CustomTestCase,
 )
 
-register_cuda_ci(est_time=387, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=387, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=261, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/cpu/test_activation.py b/test/registered/cpu/test_activation.py
index fe020c887..70843a9dc 100644
--- a/test/registered/cpu/test_activation.py
+++ b/test/registered/cpu/test_activation.py
@@ -8,7 +8,7 @@ from sglang.srt.server_args import ServerArgs, set_global_server_args_for_schedu
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 torch.manual_seed(1234)
 
diff --git a/test/registered/cpu/test_binding.py b/test/registered/cpu/test_binding.py
index f623045a7..c8bb50d8d 100644
--- a/test/registered/cpu/test_binding.py
+++ b/test/registered/cpu/test_binding.py
@@ -8,7 +8,7 @@ kernel = torch.ops.sgl_kernel
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 
 class TestBinding(CustomTestCase):
diff --git a/test/registered/cpu/test_bmm.py b/test/registered/cpu/test_bmm.py
index 93257b718..eebe10343 100644
--- a/test/registered/cpu/test_bmm.py
+++ b/test/registered/cpu/test_bmm.py
@@ -10,7 +10,7 @@ from sglang.srt.layers.quantization.fp8_utils import input_to_float8
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 torch.manual_seed(1234)
 
diff --git a/test/registered/cpu/test_causal_conv1d.py b/test/registered/cpu/test_causal_conv1d.py
index f588eafa2..378be708d 100644
--- a/test/registered/cpu/test_causal_conv1d.py
+++ b/test/registered/cpu/test_causal_conv1d.py
@@ -9,7 +9,7 @@ from utils import parametrize, precision
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 causal_conv1d_weight_pack = torch.ops.sgl_kernel.causal_conv1d_weight_pack
 causal_conv1d_fwd = torch.ops.sgl_kernel.causal_conv1d_fwd_cpu
diff --git a/test/registered/cpu/test_cpu_graph.py b/test/registered/cpu/test_cpu_graph.py
index 37d70df72..ba95bc74e 100644
--- a/test/registered/cpu/test_cpu_graph.py
+++ b/test/registered/cpu/test_cpu_graph.py
@@ -21,7 +21,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 
 class TestCPUGraph(CustomTestCase):
diff --git a/test/registered/cpu/test_decode.py b/test/registered/cpu/test_decode.py
index 316c446bd..81385d3b2 100644
--- a/test/registered/cpu/test_decode.py
+++ b/test/registered/cpu/test_decode.py
@@ -6,7 +6,7 @@ from torch.nn.functional import scaled_dot_product_attention
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 torch.manual_seed(1234)
 
diff --git a/test/registered/cpu/test_extend.py b/test/registered/cpu/test_extend.py
index 573531356..2ea6e3fba 100644
--- a/test/registered/cpu/test_extend.py
+++ b/test/registered/cpu/test_extend.py
@@ -6,7 +6,7 @@ from torch.nn.functional import scaled_dot_product_attention
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 torch.manual_seed(1234)
 
diff --git a/test/registered/cpu/test_flash_attn.py b/test/registered/cpu/test_flash_attn.py
index f7f47c454..7d79adefb 100644
--- a/test/registered/cpu/test_flash_attn.py
+++ b/test/registered/cpu/test_flash_attn.py
@@ -8,7 +8,7 @@ from utils import parametrize, precision
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 flash_attn_varlen_func = torch.ops.sgl_kernel.flash_attn_varlen_func
 
diff --git a/test/registered/cpu/test_gemm.py b/test/registered/cpu/test_gemm.py
index c5ea056e7..909e0ab27 100644
--- a/test/registered/cpu/test_gemm.py
+++ b/test/registered/cpu/test_gemm.py
@@ -16,7 +16,7 @@ from utils import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 torch.manual_seed(1234)
 
diff --git a/test/registered/cpu/test_intel_amx_attention_backend_a.py b/test/registered/cpu/test_intel_amx_attention_backend_a.py
index ba60527d2..f15afacce 100644
--- a/test/registered/cpu/test_intel_amx_attention_backend_a.py
+++ b/test/registered/cpu/test_intel_amx_attention_backend_a.py
@@ -20,7 +20,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 
 class TestIntelAMXAttnBackend(CustomTestCase):
diff --git a/test/registered/cpu/test_intel_amx_attention_backend_b.py b/test/registered/cpu/test_intel_amx_attention_backend_b.py
index 104f417c5..9b1db54ad 100644
--- a/test/registered/cpu/test_intel_amx_attention_backend_b.py
+++ b/test/registered/cpu/test_intel_amx_attention_backend_b.py
@@ -14,7 +14,7 @@ from sglang.test.test_utils import (
     intel_amx_benchmark,
 )
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 
 class TestIntelAMXAttnBackendQuant(CustomTestCase):
diff --git a/test/registered/cpu/test_intel_amx_attention_backend_c.py b/test/registered/cpu/test_intel_amx_attention_backend_c.py
index 391d21a2b..bf528df45 100644
--- a/test/registered/cpu/test_intel_amx_attention_backend_c.py
+++ b/test/registered/cpu/test_intel_amx_attention_backend_c.py
@@ -14,7 +14,7 @@ from sglang.test.test_utils import (
     intel_amx_benchmark,
 )
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 
 class TestIntelAMXAttnBackendQuant(CustomTestCase):
diff --git a/test/registered/cpu/test_mamba.py b/test/registered/cpu/test_mamba.py
index 25275e908..b5293ce82 100644
--- a/test/registered/cpu/test_mamba.py
+++ b/test/registered/cpu/test_mamba.py
@@ -8,7 +8,7 @@ from utils import precision
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 torch.manual_seed(1234)
 
diff --git a/test/registered/cpu/test_mla.py b/test/registered/cpu/test_mla.py
index f6714ced5..93867ea79 100644
--- a/test/registered/cpu/test_mla.py
+++ b/test/registered/cpu/test_mla.py
@@ -7,7 +7,7 @@ from utils import precision
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 torch.manual_seed(1234)
 
diff --git a/test/registered/cpu/test_moe.py b/test/registered/cpu/test_moe.py
index f3c817bec..7ad54d891 100644
--- a/test/registered/cpu/test_moe.py
+++ b/test/registered/cpu/test_moe.py
@@ -28,7 +28,7 @@ from utils import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 
 def fused_moe(a, w1, w2, score, topk, renormalize, prepack):
diff --git a/test/registered/cpu/test_norm.py b/test/registered/cpu/test_norm.py
index 2da4d5117..87e525b02 100644
--- a/test/registered/cpu/test_norm.py
+++ b/test/registered/cpu/test_norm.py
@@ -8,7 +8,7 @@ from utils import make_non_contiguous, parametrize, precision
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 torch.manual_seed(1234)
 
diff --git a/test/registered/cpu/test_qkv_proj_with_rope.py b/test/registered/cpu/test_qkv_proj_with_rope.py
index 2fa291860..fc21b6651 100644
--- a/test/registered/cpu/test_qkv_proj_with_rope.py
+++ b/test/registered/cpu/test_qkv_proj_with_rope.py
@@ -13,7 +13,7 @@ from sglang.srt.layers.rotary_embedding.utils import apply_rotary_emb
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 convert_weight_packed = torch.ops.sgl_kernel.convert_weight_packed
 qkv_proj_with_rope = torch.ops.sgl_kernel.qkv_proj_with_rope
diff --git a/test/registered/cpu/test_qwen3.py b/test/registered/cpu/test_qwen3.py
index e8edfeb64..21c62a550 100644
--- a/test/registered/cpu/test_qwen3.py
+++ b/test/registered/cpu/test_qwen3.py
@@ -6,7 +6,7 @@ from utils import precision
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 torch.manual_seed(1234)
 
diff --git a/test/registered/cpu/test_rope.py b/test/registered/cpu/test_rope.py
index 7246214d2..0d90f6ac1 100644
--- a/test/registered/cpu/test_rope.py
+++ b/test/registered/cpu/test_rope.py
@@ -15,7 +15,7 @@ from sglang.srt.server_args import ServerArgs, set_global_server_args_for_schedu
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 torch.manual_seed(1234)
 
diff --git a/test/registered/cpu/test_server_args_backend.py b/test/registered/cpu/test_server_args_backend.py
index 22448a7f4..9998f8f4c 100644
--- a/test/registered/cpu/test_server_args_backend.py
+++ b/test/registered/cpu/test_server_args_backend.py
@@ -4,7 +4,7 @@ from unittest.mock import patch
 from sglang.srt.server_args import ServerArgs
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 
 class TestServerArgsCPUBackend(unittest.TestCase):
diff --git a/test/registered/cpu/test_shared_expert.py b/test/registered/cpu/test_shared_expert.py
index 18b3cef78..e09444957 100644
--- a/test/registered/cpu/test_shared_expert.py
+++ b/test/registered/cpu/test_shared_expert.py
@@ -19,7 +19,7 @@ from utils import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 torch.manual_seed(1234)
 
diff --git a/test/registered/cpu/test_topk.py b/test/registered/cpu/test_topk.py
index 2d6d2bbbe..58c334832 100644
--- a/test/registered/cpu/test_topk.py
+++ b/test/registered/cpu/test_topk.py
@@ -11,7 +11,7 @@ from sglang.srt.models.llama4 import Llama4MoE
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-b-test-cpu")
+register_cpu_ci(est_time=10, suite="base-b-test-cpu")
 
 torch.manual_seed(1234)
 
diff --git a/test/registered/debug_utils/comparator/aligner/entrypoint/test_executor.py b/test/registered/debug_utils/comparator/aligner/entrypoint/test_executor.py
index 0c807c1e6..67b3a46c7 100644
--- a/test/registered/debug_utils/comparator/aligner/entrypoint/test_executor.py
+++ b/test/registered/debug_utils/comparator/aligner/entrypoint/test_executor.py
@@ -28,7 +28,7 @@ from sglang.srt.debug_utils.comparator.dims_spec import ParallelAxis, TokenLayou
 from sglang.srt.debug_utils.comparator.utils import Pair
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=15, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=15, suite="base-a-test-cpu", nightly=True)
 
 
 class TestExecuteSubPlans:
diff --git a/test/registered/debug_utils/comparator/aligner/entrypoint/test_planner.py b/test/registered/debug_utils/comparator/aligner/entrypoint/test_planner.py
index b7b9f89d7..5cf4917d6 100644
--- a/test/registered/debug_utils/comparator/aligner/entrypoint/test_planner.py
+++ b/test/registered/debug_utils/comparator/aligner/entrypoint/test_planner.py
@@ -25,7 +25,7 @@ from sglang.srt.debug_utils.comparator.dims_spec import TokenLayout
 from sglang.srt.debug_utils.comparator.utils import Pair
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=15, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=15, suite="base-a-test-cpu", nightly=True)
 
 
 def _make_meta(
diff --git a/test/registered/debug_utils/comparator/aligner/reorderer/test_executor.py b/test/registered/debug_utils/comparator/aligner/reorderer/test_executor.py
index e5b166232..568f369d0 100644
--- a/test/registered/debug_utils/comparator/aligner/reorderer/test_executor.py
+++ b/test/registered/debug_utils/comparator/aligner/reorderer/test_executor.py
@@ -22,7 +22,7 @@ from sglang.srt.debug_utils.comparator.aligner.unsharder.types import (
 from sglang.srt.debug_utils.comparator.dims_spec import ParallelAxis
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 
 def _zigzag_order(cp_size: int) -> list[int]:
diff --git a/test/registered/debug_utils/comparator/aligner/reorderer/test_planner.py b/test/registered/debug_utils/comparator/aligner/reorderer/test_planner.py
index 728df373d..18420fd80 100644
--- a/test/registered/debug_utils/comparator/aligner/reorderer/test_planner.py
+++ b/test/registered/debug_utils/comparator/aligner/reorderer/test_planner.py
@@ -24,7 +24,7 @@ from sglang.srt.debug_utils.comparator.dims_spec import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 
 class TestComputeReordererPlans:
diff --git a/test/registered/debug_utils/comparator/aligner/test_axis_aligner.py b/test/registered/debug_utils/comparator/aligner/test_axis_aligner.py
index df7591b81..36a9b624d 100644
--- a/test/registered/debug_utils/comparator/aligner/test_axis_aligner.py
+++ b/test/registered/debug_utils/comparator/aligner/test_axis_aligner.py
@@ -13,7 +13,7 @@ from sglang.srt.debug_utils.comparator.log_sink import log_sink
 from sglang.srt.debug_utils.comparator.utils import Pair
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=15, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=15, suite="base-a-test-cpu", nightly=True)
 
 
 class TestComputeAxisAlignerPlan:
diff --git a/test/registered/debug_utils/comparator/aligner/token_aligner/test_aux_loader.py b/test/registered/debug_utils/comparator/aligner/token_aligner/test_aux_loader.py
index 9e33d0262..cfecd18f2 100644
--- a/test/registered/debug_utils/comparator/aligner/token_aligner/test_aux_loader.py
+++ b/test/registered/debug_utils/comparator/aligner/token_aligner/test_aux_loader.py
@@ -19,7 +19,7 @@ from sglang.srt.debug_utils.comparator.log_sink import LogSink
 from sglang.srt.debug_utils.comparator.output_types import ErrorLog, InfoLog
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=15, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=15, suite="base-a-test-cpu", nightly=True)
 
 _sglang_plugin = _SGLangPlugin()
 _megatron_plugin = _MegatronPlugin()
diff --git a/test/registered/debug_utils/comparator/aligner/token_aligner/test_aux_plugins.py b/test/registered/debug_utils/comparator/aligner/token_aligner/test_aux_plugins.py
index 9b8cac698..1d5df4503 100644
--- a/test/registered/debug_utils/comparator/aligner/token_aligner/test_aux_plugins.py
+++ b/test/registered/debug_utils/comparator/aligner/token_aligner/test_aux_plugins.py
@@ -16,7 +16,7 @@ from sglang.srt.debug_utils.comparator.aligner.token_aligner.smart.types import
 from sglang.srt.debug_utils.comparator.dims_spec import TokenLayout
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=15, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=15, suite="base-a-test-cpu", nightly=True)
 
 _sglang_plugin = _SGLangPlugin()
 _megatron_plugin = _MegatronPlugin()
diff --git a/test/registered/debug_utils/comparator/aligner/token_aligner/test_concat_steps.py b/test/registered/debug_utils/comparator/aligner/token_aligner/test_concat_steps.py
index 26b914e20..2e7ae5e33 100644
--- a/test/registered/debug_utils/comparator/aligner/token_aligner/test_concat_steps.py
+++ b/test/registered/debug_utils/comparator/aligner/token_aligner/test_concat_steps.py
@@ -9,7 +9,7 @@ from sglang.srt.debug_utils.comparator.aligner.token_aligner.concat_steps import
 from sglang.srt.debug_utils.comparator.utils import Pair
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=15, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=15, suite="base-a-test-cpu", nightly=True)
 
 
 class TestExecuteConcat:
diff --git a/test/registered/debug_utils/comparator/aligner/token_aligner/test_executor.py b/test/registered/debug_utils/comparator/aligner/token_aligner/test_executor.py
index 41a2f6685..b71a3823e 100644
--- a/test/registered/debug_utils/comparator/aligner/token_aligner/test_executor.py
+++ b/test/registered/debug_utils/comparator/aligner/token_aligner/test_executor.py
@@ -25,7 +25,7 @@ from sglang.srt.debug_utils.comparator.dims_spec import TokenLayout
 from sglang.srt.debug_utils.comparator.utils import Pair
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=15, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=15, suite="base-a-test-cpu", nightly=True)
 
 
 def _named(tensor: torch.Tensor, names: list[str]) -> torch.Tensor:
diff --git a/test/registered/debug_utils/comparator/aligner/token_aligner/test_planner.py b/test/registered/debug_utils/comparator/aligner/token_aligner/test_planner.py
index ce02794bf..7eb13615c 100644
--- a/test/registered/debug_utils/comparator/aligner/token_aligner/test_planner.py
+++ b/test/registered/debug_utils/comparator/aligner/token_aligner/test_planner.py
@@ -23,7 +23,7 @@ from sglang.srt.debug_utils.comparator.dims_spec import TokenLayout
 from sglang.srt.debug_utils.comparator.utils import Pair
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=30, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=30, suite="base-a-test-cpu", nightly=True)
 
 
 class TestBuildTokenIndexSGLangThd:
diff --git a/test/registered/debug_utils/comparator/aligner/token_aligner/test_thd_seq_lens_loader.py b/test/registered/debug_utils/comparator/aligner/token_aligner/test_thd_seq_lens_loader.py
index 56dc7c03d..b5f348cf2 100644
--- a/test/registered/debug_utils/comparator/aligner/token_aligner/test_thd_seq_lens_loader.py
+++ b/test/registered/debug_utils/comparator/aligner/token_aligner/test_thd_seq_lens_loader.py
@@ -14,7 +14,7 @@ from sglang.srt.debug_utils.comparator.aligner.token_aligner.smart.aux_plugins i
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=15, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=15, suite="base-a-test-cpu", nightly=True)
 
 
 def _save_pt(
diff --git a/test/registered/debug_utils/comparator/aligner/unsharder/test_executor.py b/test/registered/debug_utils/comparator/aligner/unsharder/test_executor.py
index 4657b8f09..3dea0e887 100644
--- a/test/registered/debug_utils/comparator/aligner/unsharder/test_executor.py
+++ b/test/registered/debug_utils/comparator/aligner/unsharder/test_executor.py
@@ -27,7 +27,7 @@ from sglang.srt.debug_utils.comparator.dims_spec import (
 from sglang.srt.debug_utils.comparator.output_types import ReplicatedCheckResult
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 
 def _name_tensors(
diff --git a/test/registered/debug_utils/comparator/aligner/unsharder/test_parallel_info.py b/test/registered/debug_utils/comparator/aligner/unsharder/test_parallel_info.py
index f7553ef1b..0b359af8a 100644
--- a/test/registered/debug_utils/comparator/aligner/unsharder/test_parallel_info.py
+++ b/test/registered/debug_utils/comparator/aligner/unsharder/test_parallel_info.py
@@ -9,7 +9,7 @@ from sglang.srt.debug_utils.comparator.aligner.unsharder.types import AxisInfo
 from sglang.srt.debug_utils.comparator.dims_spec import ParallelAxis
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 
 class TestNormalizeParallelInfo:
diff --git a/test/registered/debug_utils/comparator/aligner/unsharder/test_planner.py b/test/registered/debug_utils/comparator/aligner/unsharder/test_planner.py
index cdf55f980..9bd14465f 100644
--- a/test/registered/debug_utils/comparator/aligner/unsharder/test_planner.py
+++ b/test/registered/debug_utils/comparator/aligner/unsharder/test_planner.py
@@ -19,7 +19,7 @@ from sglang.srt.debug_utils.comparator.aligner.unsharder.types import (
 from sglang.srt.debug_utils.comparator.dims_spec import ParallelAxis, parse_dims
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 
 class TestComputeUnsharderPlan:
diff --git a/test/registered/debug_utils/comparator/dims_spec/test_dim_parser.py b/test/registered/debug_utils/comparator/dims_spec/test_dim_parser.py
index caca5205e..caddbd698 100644
--- a/test/registered/debug_utils/comparator/dims_spec/test_dim_parser.py
+++ b/test/registered/debug_utils/comparator/dims_spec/test_dim_parser.py
@@ -12,7 +12,7 @@ from sglang.srt.debug_utils.comparator.dims_spec import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=5, suite="base-a-test-cpu", nightly=True)
 
 
 class TestParseDim:
diff --git a/test/registered/debug_utils/comparator/dims_spec/test_dims_parser.py b/test/registered/debug_utils/comparator/dims_spec/test_dims_parser.py
index b4436f4e1..4aa05f99d 100644
--- a/test/registered/debug_utils/comparator/dims_spec/test_dims_parser.py
+++ b/test/registered/debug_utils/comparator/dims_spec/test_dims_parser.py
@@ -15,7 +15,7 @@ from sglang.srt.debug_utils.comparator.dims_spec import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=5, suite="base-a-test-cpu", nightly=True)
 
 
 class TestSingletonDimUtilFilterOut:
diff --git a/test/registered/debug_utils/comparator/dims_spec/test_tensor_naming.py b/test/registered/debug_utils/comparator/dims_spec/test_tensor_naming.py
index 761172505..c8969813c 100644
--- a/test/registered/debug_utils/comparator/dims_spec/test_tensor_naming.py
+++ b/test/registered/debug_utils/comparator/dims_spec/test_tensor_naming.py
@@ -13,7 +13,7 @@ from sglang.srt.debug_utils.comparator.dims_spec import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=5, suite="base-a-test-cpu", nightly=True)
 
 
 class TestFindDimIndex:
diff --git a/test/registered/debug_utils/comparator/dims_spec/test_types.py b/test/registered/debug_utils/comparator/dims_spec/test_types.py
index 5e799ae9d..30289eed4 100644
--- a/test/registered/debug_utils/comparator/dims_spec/test_types.py
+++ b/test/registered/debug_utils/comparator/dims_spec/test_types.py
@@ -9,7 +9,7 @@ from sglang.srt.debug_utils.comparator.dims_spec import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=5, suite="base-a-test-cpu", nightly=True)
 
 
 class TestDimConstants:
diff --git a/test/registered/debug_utils/comparator/tensor_comparator/test_comparator.py b/test/registered/debug_utils/comparator/tensor_comparator/test_comparator.py
index 6e7bc6e68..754f42bba 100644
--- a/test/registered/debug_utils/comparator/tensor_comparator/test_comparator.py
+++ b/test/registered/debug_utils/comparator/tensor_comparator/test_comparator.py
@@ -14,7 +14,7 @@ from sglang.srt.debug_utils.comparator.tensor_comparator.comparator import (
 from sglang.srt.debug_utils.comparator.tensor_comparator.types import DiffInfo
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=20, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=20, suite="base-a-test-cpu", nightly=True)
 
 
 class TestComputeTensorInfo:
diff --git a/test/registered/debug_utils/comparator/tensor_comparator/test_formatter.py b/test/registered/debug_utils/comparator/tensor_comparator/test_formatter.py
index 83f95852d..342cc9f4a 100644
--- a/test/registered/debug_utils/comparator/tensor_comparator/test_formatter.py
+++ b/test/registered/debug_utils/comparator/tensor_comparator/test_formatter.py
@@ -58,7 +58,7 @@ from sglang.srt.debug_utils.comparator.tensor_comparator.types import (
 from sglang.srt.debug_utils.comparator.utils import Pair
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 _DEFAULT_PERCENTILE_LINES: list[str] = [
     "      [blue]p1        [/]    -1.8000      -1.8000   [dim]+0.00e+00[/]",
diff --git a/test/registered/debug_utils/comparator/tensor_comparator/test_types.py b/test/registered/debug_utils/comparator/tensor_comparator/test_types.py
index 6b43456e2..a7c9df2b2 100644
--- a/test/registered/debug_utils/comparator/tensor_comparator/test_types.py
+++ b/test/registered/debug_utils/comparator/tensor_comparator/test_types.py
@@ -21,7 +21,7 @@ from sglang.srt.debug_utils.comparator.tensor_comparator.types import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 
 def _make_stats(**overrides) -> TensorStats:
diff --git a/test/registered/debug_utils/comparator/test_bundle_comparator.py b/test/registered/debug_utils/comparator/test_bundle_comparator.py
index 901f885a8..64e1ae041 100644
--- a/test/registered/debug_utils/comparator/test_bundle_comparator.py
+++ b/test/registered/debug_utils/comparator/test_bundle_comparator.py
@@ -15,7 +15,7 @@ from sglang.srt.debug_utils.comparator.utils import Pair
 from sglang.srt.debug_utils.dump_loader import ValueWithMeta
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=15, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=15, suite="base-a-test-cpu", nightly=True)
 
 
 def _save_tensor(
diff --git a/test/registered/debug_utils/comparator/test_bundle_matcher.py b/test/registered/debug_utils/comparator/test_bundle_matcher.py
index 248d52d23..586e8d6e0 100644
--- a/test/registered/debug_utils/comparator/test_bundle_matcher.py
+++ b/test/registered/debug_utils/comparator/test_bundle_matcher.py
@@ -13,7 +13,7 @@ from sglang.srt.debug_utils.comparator.bundle_matcher import (
 from sglang.srt.debug_utils.comparator.utils import Pair
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=15, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=15, suite="base-a-test-cpu", nightly=True)
 
 
 def _make_row(
diff --git a/test/registered/debug_utils/comparator/test_display.py b/test/registered/debug_utils/comparator/test_display.py
index c3789bb9c..1227dcb41 100644
--- a/test/registered/debug_utils/comparator/test_display.py
+++ b/test/registered/debug_utils/comparator/test_display.py
@@ -21,7 +21,7 @@ from sglang.srt.debug_utils.comparator.output_types import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 
 def _render_rich(renderable: object) -> str:
diff --git a/test/registered/debug_utils/comparator/test_dp_utils.py b/test/registered/debug_utils/comparator/test_dp_utils.py
index db32246ba..70d484e60 100644
--- a/test/registered/debug_utils/comparator/test_dp_utils.py
+++ b/test/registered/debug_utils/comparator/test_dp_utils.py
@@ -12,7 +12,7 @@ from sglang.srt.debug_utils.comparator.dp_utils import (
 from sglang.srt.debug_utils.dump_loader import ValueWithMeta
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=15, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=15, suite="base-a-test-cpu", nightly=True)
 
 
 def _make_sglang_meta(
diff --git a/test/registered/debug_utils/comparator/test_dump_loader.py b/test/registered/debug_utils/comparator/test_dump_loader.py
index 26a68d066..9b5a25bce 100644
--- a/test/registered/debug_utils/comparator/test_dump_loader.py
+++ b/test/registered/debug_utils/comparator/test_dump_loader.py
@@ -7,7 +7,7 @@ import torch
 from sglang.srt.debug_utils.dump_loader import read_tokenizer_path
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 
 def _save_pt(
diff --git a/test/registered/debug_utils/comparator/test_entrypoint.py b/test/registered/debug_utils/comparator/test_entrypoint.py
index eb7239e13..927a218ca 100644
--- a/test/registered/debug_utils/comparator/test_entrypoint.py
+++ b/test/registered/debug_utils/comparator/test_entrypoint.py
@@ -30,7 +30,7 @@ from sglang.srt.debug_utils.comparator.output_types import (
 from sglang.srt.debug_utils.dumper import DumperConfig, _Dumper, _RecomputeStatus
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=30, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=30, suite="base-a-test-cpu", nightly=True)
 
 _FIXED_EXP_NAME = "my_exp_name"
 
diff --git a/test/registered/debug_utils/comparator/test_log_sink.py b/test/registered/debug_utils/comparator/test_log_sink.py
index 90a960d93..d825528f8 100644
--- a/test/registered/debug_utils/comparator/test_log_sink.py
+++ b/test/registered/debug_utils/comparator/test_log_sink.py
@@ -11,7 +11,7 @@ from sglang.srt.debug_utils.comparator.output_types import (
 from sglang.srt.debug_utils.comparator.report_sink import report_sink
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 
 def _make_error_log(**overrides) -> ErrorLog:
diff --git a/test/registered/debug_utils/comparator/test_manually_verify.py b/test/registered/debug_utils/comparator/test_manually_verify.py
index d69c939d7..b255aed0d 100644
--- a/test/registered/debug_utils/comparator/test_manually_verify.py
+++ b/test/registered/debug_utils/comparator/test_manually_verify.py
@@ -23,7 +23,7 @@ import torch
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=60, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=60, suite="base-a-test-cpu", nightly=True)
 
 _PUBLISH_DIR: Path = Path("/tmp/comparator_manual_verify")
 _PNG_MAGIC: bytes = b"\x89PNG"
diff --git a/test/registered/debug_utils/comparator/test_meta_overrider.py b/test/registered/debug_utils/comparator/test_meta_overrider.py
index fe2746f8c..517f8d5ce 100644
--- a/test/registered/debug_utils/comparator/test_meta_overrider.py
+++ b/test/registered/debug_utils/comparator/test_meta_overrider.py
@@ -16,7 +16,7 @@ from sglang.srt.debug_utils.comparator.meta_overrider import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 
 # ───────────────────── Unit: MetaOverrideRule ─────────────────────
diff --git a/test/registered/debug_utils/comparator/test_model_validation.py b/test/registered/debug_utils/comparator/test_model_validation.py
index cb88f72f9..cbbe9d289 100644
--- a/test/registered/debug_utils/comparator/test_model_validation.py
+++ b/test/registered/debug_utils/comparator/test_model_validation.py
@@ -44,7 +44,7 @@ from sglang.srt.debug_utils.comparator.tensor_comparator.types import (
 from sglang.srt.debug_utils.comparator.utils import Pair, _check_equal_lengths
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 
 class TestCheckEqualLengths:
diff --git a/test/registered/debug_utils/comparator/test_output_types.py b/test/registered/debug_utils/comparator/test_output_types.py
index 2ddbca1ed..8623b7c43 100644
--- a/test/registered/debug_utils/comparator/test_output_types.py
+++ b/test/registered/debug_utils/comparator/test_output_types.py
@@ -55,7 +55,7 @@ from sglang.srt.debug_utils.comparator.output_types import (
 from sglang.srt.debug_utils.comparator.utils import Pair
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 
 def _render_rich(renderable: object) -> str:
diff --git a/test/registered/debug_utils/comparator/test_per_token_visualizer.py b/test/registered/debug_utils/comparator/test_per_token_visualizer.py
index a12ef1211..f79524313 100644
--- a/test/registered/debug_utils/comparator/test_per_token_visualizer.py
+++ b/test/registered/debug_utils/comparator/test_per_token_visualizer.py
@@ -15,7 +15,7 @@ from sglang.srt.debug_utils.comparator.tensor_comparator.comparator import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=30, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=30, suite="base-a-test-cpu", nightly=True)
 
 _PNG_MAGIC: bytes = b"\x89PNG"
 
diff --git a/test/registered/debug_utils/comparator/test_preset.py b/test/registered/debug_utils/comparator/test_preset.py
index 4ccb84647..b7308b840 100644
--- a/test/registered/debug_utils/comparator/test_preset.py
+++ b/test/registered/debug_utils/comparator/test_preset.py
@@ -3,7 +3,7 @@ import pytest
 from sglang.srt.debug_utils.comparator.preset import PRESETS, expand_preset
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=5, suite="base-a-test-cpu", nightly=True)
 
 
 class TestExpandPreset:
diff --git a/test/registered/debug_utils/comparator/test_utils.py b/test/registered/debug_utils/comparator/test_utils.py
index df2677cab..9e43c70ae 100644
--- a/test/registered/debug_utils/comparator/test_utils.py
+++ b/test/registered/debug_utils/comparator/test_utils.py
@@ -17,7 +17,7 @@ from sglang.srt.debug_utils.comparator.utils import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 
 class TestCalcRelDiff:
diff --git a/test/registered/debug_utils/comparator/test_visualizer.py b/test/registered/debug_utils/comparator/test_visualizer.py
index c6aba96a9..772d26477 100644
--- a/test/registered/debug_utils/comparator/test_visualizer.py
+++ b/test/registered/debug_utils/comparator/test_visualizer.py
@@ -10,7 +10,7 @@ from sglang.srt.debug_utils.comparator.visualizer.preprocessing import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=30, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=30, suite="base-a-test-cpu", nightly=True)
 
 
 class TestPreprocessTensor:
diff --git a/test/registered/debug_utils/comparator/testing_helpers.py b/test/registered/debug_utils/comparator/testing_helpers.py
index c2e20c679..f1d177ff8 100644
--- a/test/registered/debug_utils/comparator/testing_helpers.py
+++ b/test/registered/debug_utils/comparator/testing_helpers.py
@@ -12,7 +12,7 @@ from sglang.test.ci.ci_register import register_cpu_ci
 
 register_cpu_ci(
     est_time=0,
-    suite="stage-a-test-cpu",
+    suite="base-a-test-cpu",
     nightly=True,
     disabled="helper module, no tests",
 )
diff --git a/test/registered/debug_utils/source_patcher/test_code_patcher.py b/test/registered/debug_utils/source_patcher/test_code_patcher.py
index 2106ad6a1..2e33ee7fb 100644
--- a/test/registered/debug_utils/source_patcher/test_code_patcher.py
+++ b/test/registered/debug_utils/source_patcher/test_code_patcher.py
@@ -10,7 +10,7 @@ from sglang.srt.debug_utils.source_patcher.code_patcher import (
 from sglang.srt.debug_utils.source_patcher.types import EditSpec, PatchSpec
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 SAMPLE_MODULE_NAME = "_source_patcher_test_fixtures.sample_module"
 
diff --git a/test/registered/debug_utils/source_patcher/test_dumper_integration.py b/test/registered/debug_utils/source_patcher/test_dumper_integration.py
index 2b8be98bf..ef5872979 100644
--- a/test/registered/debug_utils/source_patcher/test_dumper_integration.py
+++ b/test/registered/debug_utils/source_patcher/test_dumper_integration.py
@@ -8,7 +8,7 @@ import yaml
 from sglang.srt.debug_utils.dumper import DumperConfig, _Dumper
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 SAMPLE_MODULE_NAME = "_source_patcher_test_fixtures.sample_module"
 
diff --git a/test/registered/debug_utils/source_patcher/test_source_editor.py b/test/registered/debug_utils/source_patcher/test_source_editor.py
index fd5334be4..69d8d4951 100644
--- a/test/registered/debug_utils/source_patcher/test_source_editor.py
+++ b/test/registered/debug_utils/source_patcher/test_source_editor.py
@@ -5,7 +5,7 @@ from sglang.srt.debug_utils.source_patcher.source_editor import apply_edits
 from sglang.srt.debug_utils.source_patcher.types import EditSpec, PatchApplicationError
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=10, suite="base-a-test-cpu", nightly=True)
 
 
 class TestApplyEdits:
diff --git a/test/registered/debug_utils/test_cuda_coredump.py b/test/registered/debug_utils/test_cuda_coredump.py
index 21e2becef..d397dbe23 100644
--- a/test/registered/debug_utils/test_cuda_coredump.py
+++ b/test/registered/debug_utils/test_cuda_coredump.py
@@ -12,7 +12,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(
     est_time=10,
-    stage="stage-a",
+    stage="base-a",
     runner_config="1-gpu-small",
     disabled="Manual only: triggers intentional CUDA crash for coredump verification",
 )
diff --git a/test/registered/debug_utils/test_dump_comparator.py b/test/registered/debug_utils/test_dump_comparator.py
index 5d4945189..d269bee48 100644
--- a/test/registered/debug_utils/test_dump_comparator.py
+++ b/test/registered/debug_utils/test_dump_comparator.py
@@ -9,7 +9,7 @@ from sglang.srt.debug_utils.dump_comparator import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=30, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=30, suite="base-a-test-cpu", nightly=True)
 
 
 # ----------------------------- Unit tests -----------------------------
diff --git a/test/registered/debug_utils/test_dump_loader.py b/test/registered/debug_utils/test_dump_loader.py
index fed3a5e8a..bc6a73bfc 100644
--- a/test/registered/debug_utils/test_dump_loader.py
+++ b/test/registered/debug_utils/test_dump_loader.py
@@ -15,7 +15,7 @@ from sglang.srt.debug_utils.dump_loader import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=30, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=30, suite="base-a-test-cpu", nightly=True)
 
 
 class TestReadMeta:
diff --git a/test/registered/debug_utils/test_schedule_simulator.py b/test/registered/debug_utils/test_schedule_simulator.py
index dc7dc35c9..be82ed38c 100644
--- a/test/registered/debug_utils/test_schedule_simulator.py
+++ b/test/registered/debug_utils/test_schedule_simulator.py
@@ -25,7 +25,7 @@ from sglang.srt.debug_utils.schedule_simulator import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=120, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=120, suite="base-a-test-cpu", nightly=True)
 
 
 # ==================== Non-E2E Tests ====================
diff --git a/test/registered/debug_utils/test_tensor_dump_forward_hook.py b/test/registered/debug_utils/test_tensor_dump_forward_hook.py
index df8f7527f..4ebbfd17d 100644
--- a/test/registered/debug_utils/test_tensor_dump_forward_hook.py
+++ b/test/registered/debug_utils/test_tensor_dump_forward_hook.py
@@ -19,7 +19,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 register_cuda_ci(
     est_time=9,
-    stage="stage-b",
+    stage="base-b",
     runner_config="1-gpu-small",
     disabled="Test uses pytest-style function without TestCase class - see #17145",
 )
diff --git a/test/registered/disaggregation/test_disaggregation_basic.py b/test/registered/disaggregation/test_disaggregation_basic.py
index f14cc1756..f2ee9de56 100644
--- a/test/registered/disaggregation/test_disaggregation_basic.py
+++ b/test/registered/disaggregation/test_disaggregation_basic.py
@@ -21,7 +21,7 @@ from sglang.test.test_utils import (
     DEFAULT_TARGET_MODEL_EAGLE3,
 )
 
-register_cuda_ci(est_time=509, stage="stage-b", runner_config="2-gpu-large")
+register_cuda_ci(est_time=509, stage="base-b", runner_config="2-gpu-large")
 
 
 class TestDisaggregationAccuracy(PauseResumeInPlaceMixin, PDDisaggregationServerBase):
diff --git a/test/registered/disaggregation/test_disaggregation_decode_offload.py b/test/registered/disaggregation/test_disaggregation_decode_offload.py
index 689175897..9a8baba68 100644
--- a/test/registered/disaggregation/test_disaggregation_decode_offload.py
+++ b/test/registered/disaggregation/test_disaggregation_decode_offload.py
@@ -19,7 +19,7 @@ from sglang.test.test_utils import (
 # Increasing estimated time since we run evaluation twice
 register_cuda_ci(
     est_time=600,
-    stage="stage-b",
+    stage="base-b",
     runner_config="2-gpu-large",
     disabled="Temporarily disable the flaky test.",
 )
diff --git a/test/registered/disaggregation/test_disaggregation_xpu.py b/test/registered/disaggregation/test_disaggregation_xpu.py
index e3a5e6a94..8034966e6 100644
--- a/test/registered/disaggregation/test_disaggregation_xpu.py
+++ b/test/registered/disaggregation/test_disaggregation_xpu.py
@@ -26,7 +26,7 @@ from sglang.test.test_utils import DEFAULT_SMALL_MODEL_NAME_FOR_TEST_QWEN
 
 register_cuda_ci(
     est_time=300,
-    stage="stage-a",
+    stage="base-a",
     runner_config="1-gpu-small",
     disabled="Intel XPU only — not available in standard CUDA CI",
 )
diff --git a/test/registered/disaggregation/test_specv2_kvcache_offloading.py b/test/registered/disaggregation/test_specv2_kvcache_offloading.py
index 8d70919d1..d9876dbcd 100644
--- a/test/registered/disaggregation/test_specv2_kvcache_offloading.py
+++ b/test/registered/disaggregation/test_specv2_kvcache_offloading.py
@@ -17,7 +17,7 @@ from sglang.srt.disaggregation.decode_kvcache_offload_manager import (
 )
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=8, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=8, stage="base-b", runner_config="1-gpu-small")
 
 
 def _make_mock_req(
diff --git a/test/registered/distributed/test_data_parallelism.py b/test/registered/distributed/test_data_parallelism.py
index f79b8eae7..acf68ed27 100644
--- a/test/registered/distributed/test_data_parallelism.py
+++ b/test/registered/distributed/test_data_parallelism.py
@@ -14,7 +14,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=91, stage="stage-b", runner_config="2-gpu-large")
+register_cuda_ci(est_time=91, stage="base-b", runner_config="2-gpu-large")
 register_amd_ci(est_time=73, suite="stage-b-test-2-gpu-large-amd")
 
 
diff --git a/test/registered/distributed/test_disaggregation_aarch64.py b/test/registered/distributed/test_disaggregation_aarch64.py
index 850343569..5c8630282 100644
--- a/test/registered/distributed/test_disaggregation_aarch64.py
+++ b/test/registered/distributed/test_disaggregation_aarch64.py
@@ -13,7 +13,7 @@ from sglang.test.test_utils import (
     popen_launch_pd_server,
 )
 
-register_cuda_ci(est_time=300, stage="stage-c", runner_config="4-gpu-gb200")
+register_cuda_ci(est_time=300, stage="base-c", runner_config="4-gpu-gb200")
 
 
 class TestDisaggregationMooncakeAARCH64Accuracy(PDDisaggregationServerBase):
diff --git a/test/registered/distributed/test_disaggregation_decode_radix_cache.py b/test/registered/distributed/test_disaggregation_decode_radix_cache.py
index 57a15bc3d..249443c58 100644
--- a/test/registered/distributed/test_disaggregation_decode_radix_cache.py
+++ b/test/registered/distributed/test_disaggregation_decode_radix_cache.py
@@ -16,7 +16,7 @@ from sglang.test.test_utils import (
     try_cached_model,
 )
 
-register_cuda_ci(est_time=300, stage="stage-c", runner_config="8-gpu-h20")
+register_cuda_ci(est_time=300, stage="base-c", runner_config="8-gpu-h20")
 
 
 def _has_nixl():
diff --git a/test/registered/distributed/test_disaggregation_different_tp.py b/test/registered/distributed/test_disaggregation_different_tp.py
index c9d725b87..e227cd73a 100644
--- a/test/registered/distributed/test_disaggregation_different_tp.py
+++ b/test/registered/distributed/test_disaggregation_different_tp.py
@@ -16,7 +16,7 @@ from sglang.test.test_utils import (
     try_cached_model,
 )
 
-register_cuda_ci(est_time=375, stage="stage-c", runner_config="8-gpu-h20")
+register_cuda_ci(est_time=375, stage="base-c", runner_config="8-gpu-h20")
 
 
 class TestDisaggregationMooncakePrefillLargerTP(PDDisaggregationServerBase):
diff --git a/test/registered/distributed/test_disaggregation_dp_attention.py b/test/registered/distributed/test_disaggregation_dp_attention.py
index b64a5559b..d7fba1c27 100644
--- a/test/registered/distributed/test_disaggregation_dp_attention.py
+++ b/test/registered/distributed/test_disaggregation_dp_attention.py
@@ -16,7 +16,7 @@ from sglang.test.test_utils import (
     try_cached_model,
 )
 
-register_cuda_ci(est_time=443, stage="stage-c", runner_config="8-gpu-h20")
+register_cuda_ci(est_time=443, stage="base-c", runner_config="8-gpu-h20")
 
 
 class TestDisaggregationDPAttention(PDDisaggregationServerBase):
diff --git a/test/registered/distributed/test_disaggregation_dsv4.py b/test/registered/distributed/test_disaggregation_dsv4.py
index 392996508..e20513fbc 100644
--- a/test/registered/distributed/test_disaggregation_dsv4.py
+++ b/test/registered/distributed/test_disaggregation_dsv4.py
@@ -12,7 +12,7 @@ from sglang.test.test_utils import (
     try_cached_model,
 )
 
-register_cuda_ci(est_time=250, stage="stage-c", runner_config="dsv4-8-gpu-h200")
+register_cuda_ci(est_time=250, stage="base-c", runner_config="dsv4-8-gpu-h200")
 
 DSV4_FLASH_MODEL = "sgl-project/DeepSeek-V4-Flash-FP8"
 
diff --git a/test/registered/distributed/test_disaggregation_pp.py b/test/registered/distributed/test_disaggregation_pp.py
index e04c7b271..05d052214 100644
--- a/test/registered/distributed/test_disaggregation_pp.py
+++ b/test/registered/distributed/test_disaggregation_pp.py
@@ -14,7 +14,7 @@ from sglang.test.test_utils import (
     try_cached_model,
 )
 
-register_cuda_ci(est_time=216, stage="stage-c", runner_config="8-gpu-h20")
+register_cuda_ci(est_time=216, stage="base-c", runner_config="8-gpu-h20")
 
 
 class TestDisaggregationPrefillPPAccuracy(PDDisaggregationServerBase):
diff --git a/test/registered/distributed/test_dp_attention.py b/test/registered/distributed/test_dp_attention.py
index cae258138..583d8fde8 100644
--- a/test/registered/distributed/test_dp_attention.py
+++ b/test/registered/distributed/test_dp_attention.py
@@ -21,7 +21,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=420, stage="stage-b", runner_config="2-gpu-large")
+register_cuda_ci(est_time=420, stage="base-b", runner_config="2-gpu-large")
 
 
 class TestDPAttentionDP2TP2(
diff --git a/test/registered/distributed/test_flashinfer_fusion_preflight.py b/test/registered/distributed/test_flashinfer_fusion_preflight.py
index 118a0b554..ea06d0124 100644
--- a/test/registered/distributed/test_flashinfer_fusion_preflight.py
+++ b/test/registered/distributed/test_flashinfer_fusion_preflight.py
@@ -11,7 +11,7 @@ from sglang.srt.utils import get_cuda_driver_bindings, is_flashinfer_available
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=30, stage="stage-b", runner_config="2-gpu-large")
+register_cuda_ci(est_time=30, stage="base-b", runner_config="2-gpu-large")
 
 WORLD_SIZE = 2
 
diff --git a/test/registered/distributed/test_parallel_state.py b/test/registered/distributed/test_parallel_state.py
index 1a3994c49..1a6959e46 100644
--- a/test/registered/distributed/test_parallel_state.py
+++ b/test/registered/distributed/test_parallel_state.py
@@ -43,7 +43,7 @@ import pytest
 
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=8, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=8, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=8, suite="stage-b-test-1-gpu-small-amd")
 
 # Import the actual parallel_state module
diff --git a/test/registered/distributed/test_pp_single_node.py b/test/registered/distributed/test_pp_single_node.py
index 519bb5573..85dafad5c 100644
--- a/test/registered/distributed/test_pp_single_node.py
+++ b/test/registered/distributed/test_pp_single_node.py
@@ -32,7 +32,7 @@ from sglang.test.test_utils import (
     run_bench_one_batch_server,
 )
 
-register_cuda_ci(est_time=554, stage="stage-c", runner_config="4-gpu-h100")
+register_cuda_ci(est_time=554, stage="base-c", runner_config="4-gpu-h100")
 register_amd_ci(est_time=650, suite="stage-c-test-4-gpu-amd")
 
 
diff --git a/test/registered/dllm/test_llada2_mini.py b/test/registered/dllm/test_llada2_mini.py
index 33ed3c861..7b869c70c 100644
--- a/test/registered/dllm/test_llada2_mini.py
+++ b/test/registered/dllm/test_llada2_mini.py
@@ -1,6 +1,6 @@
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=139, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=139, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=330, suite="stage-b-test-1-gpu-small-amd")
 
 import unittest
diff --git a/test/registered/dsa_models_e2e/test_dsa_glm5_dp_mtp.py b/test/registered/dsa_models_e2e/test_dsa_glm5_dp_mtp.py
index 7ba2b6101..b81f47495 100644
--- a/test/registered/dsa_models_e2e/test_dsa_glm5_dp_mtp.py
+++ b/test/registered/dsa_models_e2e/test_dsa_glm5_dp_mtp.py
@@ -10,7 +10,7 @@ from sglang.test.server_fixtures.dsa_mtp_fixture import (
 
 register_cuda_ci(
     est_time=400,
-    stage="stage-c",
+    stage="base-c",
     runner_config="8-gpu-h200",
 )
 
diff --git a/test/registered/dsa_models_e2e/test_dsa_glm5_tp_mtp.py b/test/registered/dsa_models_e2e/test_dsa_glm5_tp_mtp.py
index 44c7afe64..5e5415de6 100644
--- a/test/registered/dsa_models_e2e/test_dsa_glm5_tp_mtp.py
+++ b/test/registered/dsa_models_e2e/test_dsa_glm5_tp_mtp.py
@@ -10,7 +10,7 @@ from sglang.test.server_fixtures.dsa_mtp_fixture import (
 
 register_cuda_ci(
     est_time=400,
-    stage="stage-c",
+    stage="base-c",
     runner_config="8-gpu-h200",
 )
 
diff --git a/test/registered/dsv4/test_deepseek_v4_flash_fp4_b200.py b/test/registered/dsv4/test_deepseek_v4_flash_fp4_b200.py
index 6021f6109..212f2b6b2 100644
--- a/test/registered/dsv4/test_deepseek_v4_flash_fp4_b200.py
+++ b/test/registered/dsv4/test_deepseek_v4_flash_fp4_b200.py
@@ -4,7 +4,7 @@ Launches TP=4 with flashinfer_mxfp4 MoE runner + EAGLE speculative decoding.
 Runs 12 ServerSanity probes (correctness, streaming, concurrency, determinism)
 plus a GSM8K accuracy gate.
 
-Registry: stage-c-test-dsv4-4-gpu-b200 (per-commit, 4x B200)
+Registry: base-c-test-dsv4-4-gpu-b200 (per-commit, 4x B200)
 """
 
 import unittest
@@ -21,7 +21,7 @@ from sglang.test.test_utils import (
     try_cached_model,
 )
 
-register_cuda_ci(est_time=700, stage="stage-c", runner_config="dsv4-4-gpu-b200")
+register_cuda_ci(est_time=700, stage="base-c", runner_config="dsv4-4-gpu-b200")
 
 MODEL = "deepseek-ai/DeepSeek-V4-Flash"
 SERVER_LAUNCH_TIMEOUT = 3600
diff --git a/test/registered/dsv4/test_deepseek_v4_flash_fp4_h200.py b/test/registered/dsv4/test_deepseek_v4_flash_fp4_h200.py
index bc1469fe1..a7ec88931 100644
--- a/test/registered/dsv4/test_deepseek_v4_flash_fp4_h200.py
+++ b/test/registered/dsv4/test_deepseek_v4_flash_fp4_h200.py
@@ -4,7 +4,7 @@ Launches TP=4 with Marlin FP4 MoE runner + EAGLE speculative decoding.
 Runs 12 ServerSanity probes (correctness, streaming, concurrency, determinism)
 plus a GSM8K accuracy gate.
 
-Registry: stage-c-test-dsv4-8-gpu-h200 (per-commit, 8x H200 — only 4 used by TP=4)
+Registry: base-c-test-dsv4-8-gpu-h200 (per-commit, 8x H200 — only 4 used by TP=4)
 """
 
 import unittest
@@ -21,7 +21,7 @@ from sglang.test.test_utils import (
     try_cached_model,
 )
 
-register_cuda_ci(est_time=370, stage="stage-c", runner_config="dsv4-8-gpu-h200")
+register_cuda_ci(est_time=370, stage="base-c", runner_config="dsv4-8-gpu-h200")
 
 
 def _flashinfer_has_sm90_cutlass_mxfp4() -> bool:
diff --git a/test/registered/dsv4/test_deepseek_v4_flash_fp4_megamoe_b200.py b/test/registered/dsv4/test_deepseek_v4_flash_fp4_megamoe_b200.py
index ce962adfb..6cf20c8b3 100644
--- a/test/registered/dsv4/test_deepseek_v4_flash_fp4_megamoe_b200.py
+++ b/test/registered/dsv4/test_deepseek_v4_flash_fp4_megamoe_b200.py
@@ -4,7 +4,7 @@ Launches TP=4 with flashinfer_mxfp4 MoE runner + EAGLE speculative decoding.
 Runs 12 ServerSanity probes (correctness, streaming, concurrency, determinism)
 plus a GSM8K accuracy gate.
 
-Registry: stage-c-test-dsv4-4-gpu-b200 (per-commit, 4x B200)
+Registry: base-c-test-dsv4-4-gpu-b200 (per-commit, 4x B200)
 """
 
 import unittest
@@ -21,7 +21,7 @@ from sglang.test.test_utils import (
     try_cached_model,
 )
 
-register_cuda_ci(est_time=900, stage="stage-c", runner_config="dsv4-4-gpu-b200")
+register_cuda_ci(est_time=900, stage="base-c", runner_config="dsv4-4-gpu-b200")
 
 MODEL = "deepseek-ai/DeepSeek-V4-Flash"
 SERVER_LAUNCH_TIMEOUT = 3600
diff --git a/test/registered/dsv4/test_deepseek_v4_flash_fp8_h200.py b/test/registered/dsv4/test_deepseek_v4_flash_fp8_h200.py
index c22c72b3b..3fc82e4c8 100644
--- a/test/registered/dsv4/test_deepseek_v4_flash_fp8_h200.py
+++ b/test/registered/dsv4/test_deepseek_v4_flash_fp8_h200.py
@@ -5,7 +5,7 @@ with FP4 experts disabled via SGLANG_DSV4_FP4_EXPERTS=0.
 Runs 12 ServerSanity probes (correctness, streaming, concurrency, determinism)
 plus a GSM8K accuracy gate.
 
-Registry: stage-c-test-dsv4-8-gpu-h200 (per-commit, 8x H200 — only 4 used by TP=4)
+Registry: base-c-test-dsv4-8-gpu-h200 (per-commit, 8x H200 — only 4 used by TP=4)
 """
 
 import unittest
@@ -22,7 +22,7 @@ from sglang.test.test_utils import (
     try_cached_model,
 )
 
-register_cuda_ci(est_time=280, stage="stage-c", runner_config="dsv4-8-gpu-h200")
+register_cuda_ci(est_time=280, stage="base-c", runner_config="dsv4-8-gpu-h200")
 
 MODEL_FP8 = "sgl-project/DeepSeek-V4-Flash-FP8"
 SERVER_LAUNCH_TIMEOUT = 3600
diff --git a/test/registered/ep/test_deepep_small.py b/test/registered/ep/test_deepep_small.py
index abc672cfc..b5a75a006 100644
--- a/test/registered/ep/test_deepep_small.py
+++ b/test/registered/ep/test_deepep_small.py
@@ -16,7 +16,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=478, stage="stage-c", runner_config="deepep-4-gpu-h100")
+register_cuda_ci(est_time=478, stage="base-c", runner_config="deepep-4-gpu-h100")
 
 
 class TestPureDP(CustomTestCase):
diff --git a/test/registered/ep/test_mooncake_ep_small.py b/test/registered/ep/test_mooncake_ep_small.py
index 3fac6eb69..aa6b11425 100644
--- a/test/registered/ep/test_mooncake_ep_small.py
+++ b/test/registered/ep/test_mooncake_ep_small.py
@@ -15,7 +15,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=82, stage="stage-c", runner_config="deepep-4-gpu-h100")
+register_cuda_ci(est_time=82, stage="base-c", runner_config="deepep-4-gpu-h100")
 
 ib_devices = get_rdma_devices_args()
 
diff --git a/test/registered/function_call/test_kimik2_detector.py b/test/registered/function_call/test_kimik2_detector.py
index be0cd83c1..7352472cb 100644
--- a/test/registered/function_call/test_kimik2_detector.py
+++ b/test/registered/function_call/test_kimik2_detector.py
@@ -11,7 +11,7 @@ from sglang.srt.function_call.kimik2_detector import (
 from sglang.srt.parser.reasoning_parser import KimiK2Detector as KimiK2ReasoningDetector
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(5, "stage-a-test-cpu")
+register_cpu_ci(5, "base-a-test-cpu")
 
 
 def _make_tool(name, parameters=None):
diff --git a/test/registered/hicache/test_hicache_storage.py b/test/registered/hicache/test_hicache_storage.py
index 1ae355593..b4f7ae013 100644
--- a/test/registered/hicache/test_hicache_storage.py
+++ b/test/registered/hicache/test_hicache_storage.py
@@ -1,6 +1,6 @@
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=99, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=99, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=300, suite="stage-b-test-1-gpu-small-amd")
 
 import time
diff --git a/test/registered/hicache/test_hicache_storage_3fs_backend.py b/test/registered/hicache/test_hicache_storage_3fs_backend.py
index 124b43d83..aac442efd 100644
--- a/test/registered/hicache/test_hicache_storage_3fs_backend.py
+++ b/test/registered/hicache/test_hicache_storage_3fs_backend.py
@@ -13,8 +13,8 @@ from test_hicache_storage_file_backend import HiCacheStorageBaseMixin
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=150, stage="stage-b", runner_config="2-gpu-large")
-register_amd_ci(est_time=300, suite="stage-b-test-2-gpu-large")
+register_cuda_ci(est_time=150, stage="base-b", runner_config="2-gpu-large")
+register_amd_ci(est_time=300, suite="base-b-test-2-gpu-large")
 
 
 class HiCacheStorage3FSBackendBaseMixin(HiCacheStorageBaseMixin):
diff --git a/test/registered/hicache/test_hicache_storage_file_backend.py b/test/registered/hicache/test_hicache_storage_file_backend.py
index 163a9fe7b..8f38636f5 100644
--- a/test/registered/hicache/test_hicache_storage_file_backend.py
+++ b/test/registered/hicache/test_hicache_storage_file_backend.py
@@ -31,7 +31,7 @@ from sglang.test.test_utils import (
 )
 from sglang.utils import wait_for_http_ready
 
-register_cuda_ci(est_time=148, stage="stage-b", runner_config="2-gpu-large")
+register_cuda_ci(est_time=148, stage="base-b", runner_config="2-gpu-large")
 register_amd_ci(est_time=526, suite="stage-b-test-2-gpu-large-amd")
 
 
diff --git a/test/registered/hicache/test_hicache_storage_mooncake_backend.py b/test/registered/hicache/test_hicache_storage_mooncake_backend.py
index 15ae0f05b..949bfda2c 100644
--- a/test/registered/hicache/test_hicache_storage_mooncake_backend.py
+++ b/test/registered/hicache/test_hicache_storage_mooncake_backend.py
@@ -21,7 +21,7 @@ from sglang.test.test_utils import (
     is_in_ci,
 )
 
-register_cuda_ci(est_time=236, stage="stage-b", runner_config="2-gpu-large")
+register_cuda_ci(est_time=236, stage="base-b", runner_config="2-gpu-large")
 
 
 class HiCacheStorageMooncakeBackendBaseMixin(HiCacheStorageBaseMixin):
diff --git a/test/registered/hicache/test_hicache_storage_runtime_attach_detach.py b/test/registered/hicache/test_hicache_storage_runtime_attach_detach.py
index cf74ba4e8..8dbb452c3 100644
--- a/test/registered/hicache/test_hicache_storage_runtime_attach_detach.py
+++ b/test/registered/hicache/test_hicache_storage_runtime_attach_detach.py
@@ -28,7 +28,7 @@ from sglang.test.test_utils import (
 )
 from sglang.utils import wait_for_http_ready
 
-register_cuda_ci(est_time=139, stage="stage-b", runner_config="2-gpu-large")
+register_cuda_ci(est_time=139, stage="base-b", runner_config="2-gpu-large")
 
 
 class TestHiCacheStorageRuntimeAttachDetach(CustomTestCase):
diff --git a/test/registered/hicache/test_hicache_variants.py b/test/registered/hicache/test_hicache_variants.py
index 671220446..36e2c4148 100644
--- a/test/registered/hicache/test_hicache_variants.py
+++ b/test/registered/hicache/test_hicache_variants.py
@@ -1,6 +1,6 @@
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=450, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=450, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=524, suite="stage-b-test-1-gpu-small-amd")
 """
 Consolidated HiCache variant tests.
diff --git a/test/registered/input_embedding/test_input_embeddings.py b/test/registered/input_embedding/test_input_embeddings.py
index 32736e8ef..fc49562cd 100644
--- a/test/registered/input_embedding/test_input_embeddings.py
+++ b/test/registered/input_embedding/test_input_embeddings.py
@@ -16,7 +16,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=42, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=42, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=38, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/input_embedding/test_input_embeds_chunked.py b/test/registered/input_embedding/test_input_embeds_chunked.py
index 800058930..878df0136 100644
--- a/test/registered/input_embedding/test_input_embeds_chunked.py
+++ b/test/registered/input_embedding/test_input_embeds_chunked.py
@@ -30,7 +30,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=43, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=43, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=43, suite="stage-b-test-1-gpu-small-amd")
 
 CHUNKED_PREFILL_SIZE = 256
diff --git a/test/registered/kernels/test_nsa_indexer.py b/test/registered/kernels/test_nsa_indexer.py
index 56958121c..e1d877012 100644
--- a/test/registered/kernels/test_nsa_indexer.py
+++ b/test/registered/kernels/test_nsa_indexer.py
@@ -24,7 +24,7 @@ from sglang.srt.model_executor.forward_batch_info import ForwardBatch, ForwardMo
 from sglang.srt.server_args import ServerArgs, set_global_server_args_for_scheduler
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=18, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=18, stage="base-b", runner_config="1-gpu-large")
 
 # Global configuration for all indexer tests
 DEFAULT_CONFIG = {
diff --git a/test/registered/language/test_srt_backend.py b/test/registered/language/test_srt_backend.py
index 8b775bea9..f39bef459 100644
--- a/test/registered/language/test_srt_backend.py
+++ b/test/registered/language/test_srt_backend.py
@@ -20,7 +20,7 @@ from sglang.test.test_programs import (
 )
 from sglang.test.test_utils import DEFAULT_MODEL_NAME_FOR_TEST, CustomTestCase
 
-register_cuda_ci(est_time=79, stage="stage-a", runner_config="1-gpu-small")
+register_cuda_ci(est_time=79, stage="base-a", runner_config="1-gpu-small")
 register_amd_ci(est_time=120, suite="stage-a-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/layers/mamba/test_causal_conv1d.py b/test/registered/layers/mamba/test_causal_conv1d.py
index 06f0958f5..e3e8d2a9d 100644
--- a/test/registered/layers/mamba/test_causal_conv1d.py
+++ b/test/registered/layers/mamba/test_causal_conv1d.py
@@ -2,7 +2,7 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=11, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=25, suite="stage-b-test-1-gpu-small-amd")
 
 # Adapted from https://github.com/vllm-project/vllm/blob/main/tests/kernels/mamba/test_causal_conv1d.py
diff --git a/test/registered/layers/mamba/test_mamba2_mixer.py b/test/registered/layers/mamba/test_mamba2_mixer.py
index 5fd7e62d1..d5a0c823d 100644
--- a/test/registered/layers/mamba/test_mamba2_mixer.py
+++ b/test/registered/layers/mamba/test_mamba2_mixer.py
@@ -18,7 +18,7 @@ from sglang.srt.distributed.parallel_state import (
 from sglang.srt.utils import get_device, get_device_count
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=32, stage="stage-b", runner_config="2-gpu-large")
+register_cuda_ci(est_time=32, stage="base-b", runner_config="2-gpu-large")
 
 NUM_GPUS = 2
 
diff --git a/test/registered/layers/mamba/test_mamba_ssm.py b/test/registered/layers/mamba/test_mamba_ssm.py
index 510bd8aee..39c749eb4 100644
--- a/test/registered/layers/mamba/test_mamba_ssm.py
+++ b/test/registered/layers/mamba/test_mamba_ssm.py
@@ -2,7 +2,7 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=10, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=20, suite="stage-b-test-1-gpu-small-amd")
 
 # Adapted from https://github.com/vllm-project/vllm/blob/633f943e30a4444d890d26b81850f7217736f840/tests/kernels/mamba/test_mamba_ssm_ssd.py
diff --git a/test/registered/layers/mamba/test_mamba_ssm_ssd.py b/test/registered/layers/mamba/test_mamba_ssm_ssd.py
index bd2b1b255..2e9d93a4d 100644
--- a/test/registered/layers/mamba/test_mamba_ssm_ssd.py
+++ b/test/registered/layers/mamba/test_mamba_ssm_ssd.py
@@ -2,7 +2,7 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=10, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=34, suite="stage-b-test-1-gpu-small-amd")
 
 # Adapted from https://github.com/vllm-project/vllm/blob/633f943e30a4444d890d26b81850f7217736f840/tests/kernels/mamba/test_mamba_ssm_ssd.py
diff --git a/test/registered/layers/test_fla_layernorm_guard.py b/test/registered/layers/test_fla_layernorm_guard.py
index b65415f7b..d860d3b71 100644
--- a/test/registered/layers/test_fla_layernorm_guard.py
+++ b/test/registered/layers/test_fla_layernorm_guard.py
@@ -19,7 +19,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(
     est_time=60,
-    stage="stage-b",
+    stage="base-b",
     runner_config="2-gpu-large",
     disabled="Temporarily disabled",
 )
diff --git a/test/registered/lora/test_fused_moe_lora_kernel.py b/test/registered/lora/test_fused_moe_lora_kernel.py
index 5fc4bc809..9b152464c 100644
--- a/test/registered/lora/test_fused_moe_lora_kernel.py
+++ b/test/registered/lora/test_fused_moe_lora_kernel.py
@@ -15,7 +15,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
 
 # ==============================================================================
 
-register_cuda_ci(est_time=28, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=28, stage="base-b", runner_config="1-gpu-large")
 
 
 def round_up(x, base):
diff --git a/test/registered/lora/test_lora_hf_sgl_logprob_diff.py b/test/registered/lora/test_lora_hf_sgl_logprob_diff.py
index 0f17d0788..efebc2e1b 100644
--- a/test/registered/lora/test_lora_hf_sgl_logprob_diff.py
+++ b/test/registered/lora/test_lora_hf_sgl_logprob_diff.py
@@ -41,7 +41,7 @@ from sglang.test.test_utils import DEFAULT_PORT_FOR_SRT_TEST_RUNNER, CustomTestC
 
 register_cuda_ci(
     est_time=150,
-    stage="stage-b",
+    stage="base-b",
     runner_config="1-gpu-small",
 )
 register_amd_ci(
diff --git a/test/registered/lora/test_lora_moe_vllm_sgl_logprob_diff.py b/test/registered/lora/test_lora_moe_vllm_sgl_logprob_diff.py
index a2f397a5e..a35e69dd9 100644
--- a/test/registered/lora/test_lora_moe_vllm_sgl_logprob_diff.py
+++ b/test/registered/lora/test_lora_moe_vllm_sgl_logprob_diff.py
@@ -26,7 +26,7 @@ from sglang.test.runners import SRTRunner
 
 register_cuda_ci(
     est_time=50,
-    stage="stage-b",
+    stage="base-b",
     runner_config="1-gpu-large",
 )
 
diff --git a/test/registered/lora/test_lora_overlap_loading.py b/test/registered/lora/test_lora_overlap_loading.py
index 36964f7f3..ea7765cf6 100644
--- a/test/registered/lora/test_lora_overlap_loading.py
+++ b/test/registered/lora/test_lora_overlap_loading.py
@@ -29,7 +29,7 @@ from sglang.test.lora_utils import (
 )
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=48, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=48, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=75, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/lora/test_lora_qwen3_5_35b_a3b_logprob_diff.py b/test/registered/lora/test_lora_qwen3_5_35b_a3b_logprob_diff.py
index ea959653b..489418f80 100644
--- a/test/registered/lora/test_lora_qwen3_5_35b_a3b_logprob_diff.py
+++ b/test/registered/lora/test_lora_qwen3_5_35b_a3b_logprob_diff.py
@@ -36,7 +36,7 @@ from sglang.test.test_utils import CustomTestCase
 
 register_cuda_ci(
     est_time=110,
-    stage="stage-c",
+    stage="base-c",
     runner_config="4-gpu-b200",
 )
 
diff --git a/test/registered/lora/test_lora_qwen3_vl_30b_a3b_instruct_logprob_diff.py b/test/registered/lora/test_lora_qwen3_vl_30b_a3b_instruct_logprob_diff.py
index 90cafd6a5..01a62e78b 100644
--- a/test/registered/lora/test_lora_qwen3_vl_30b_a3b_instruct_logprob_diff.py
+++ b/test/registered/lora/test_lora_qwen3_vl_30b_a3b_instruct_logprob_diff.py
@@ -36,7 +36,7 @@ from sglang.test.test_utils import CustomTestCase
 
 register_cuda_ci(
     est_time=110,
-    stage="stage-c",
+    stage="base-c",
     runner_config="4-gpu-b200",
 )
 
diff --git a/test/registered/lora/test_lora_tp.py b/test/registered/lora/test_lora_tp.py
index 7f2ad6b78..3ed859d11 100644
--- a/test/registered/lora/test_lora_tp.py
+++ b/test/registered/lora/test_lora_tp.py
@@ -31,7 +31,7 @@ from sglang.test.test_utils import CustomTestCase, is_in_ci
 
 register_cuda_ci(
     est_time=230,
-    stage="stage-c",
+    stage="base-c",
     runner_config="8-gpu-h200",
 )
 register_amd_ci(
diff --git a/test/registered/lora/test_lora_update.py b/test/registered/lora/test_lora_update.py
index ea01a512d..977817961 100644
--- a/test/registered/lora/test_lora_update.py
+++ b/test/registered/lora/test_lora_update.py
@@ -36,7 +36,7 @@ from sglang.test.test_utils import (
 
 register_cuda_ci(
     est_time=487,
-    stage="stage-b",
+    stage="base-b",
     runner_config="1-gpu-large",
 )
 
diff --git a/test/registered/lora/test_multi_lora_backend.py b/test/registered/lora/test_multi_lora_backend.py
index 9cb87da63..5f3e372e8 100644
--- a/test/registered/lora/test_multi_lora_backend.py
+++ b/test/registered/lora/test_multi_lora_backend.py
@@ -25,7 +25,7 @@ from sglang.test.lora_utils import (
 )
 from sglang.test.test_utils import CustomTestCase, is_in_ci
 
-register_cuda_ci(est_time=99, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=99, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=100, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/lora/test_virtual_experts_kernels.py b/test/registered/lora/test_virtual_experts_kernels.py
index 94106b123..91423abc8 100644
--- a/test/registered/lora/test_virtual_experts_kernels.py
+++ b/test/registered/lora/test_virtual_experts_kernels.py
@@ -30,7 +30,7 @@ import torch
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=15, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=15, stage="base-b", runner_config="1-gpu-small")
 
 from sglang.srt.lora.triton_ops.virtual_experts import (
     _align_block_size_jit,
diff --git a/test/registered/mla/test_flashmla.py b/test/registered/mla/test_flashmla.py
index a73086560..18c221c8d 100644
--- a/test/registered/mla/test_flashmla.py
+++ b/test/registered/mla/test_flashmla.py
@@ -21,7 +21,7 @@ from sglang.test.test_utils import (
 )
 
 # FlashMLA attention backend tests with MTP speculative decoding
-register_cuda_ci(est_time=160, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=160, stage="base-b", runner_config="1-gpu-large")
 
 
 class TestFlashMLAMTP(CustomTestCase):
diff --git a/test/registered/mla/test_mla_flashinfer.py b/test/registered/mla/test_mla_flashinfer.py
index e57eddbb3..5c000cc75 100644
--- a/test/registered/mla/test_mla_flashinfer.py
+++ b/test/registered/mla/test_mla_flashinfer.py
@@ -15,7 +15,7 @@ from sglang.test.test_utils import (
 )
 
 # FlashInfer MLA backend tests with MTP speculative decoding
-register_cuda_ci(est_time=130, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=130, stage="base-b", runner_config="1-gpu-large")
 
 
 class TestFlashinferMLAMTP(CustomTestCase):
diff --git a/test/registered/mla/test_mla_fp8.py b/test/registered/mla/test_mla_fp8.py
index 97ce56e38..6ef747c3b 100644
--- a/test/registered/mla/test_mla_fp8.py
+++ b/test/registered/mla/test_mla_fp8.py
@@ -13,7 +13,7 @@ from sglang.test.test_utils import (
 )
 
 # MLA FP8 KV cache test with MGSM evaluation
-register_cuda_ci(est_time=104, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=104, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=800, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/mla/test_mla_int8_deepseek_v3.py b/test/registered/mla/test_mla_int8_deepseek_v3.py
index 6d3136834..46f9a8e1e 100644
--- a/test/registered/mla/test_mla_int8_deepseek_v3.py
+++ b/test/registered/mla/test_mla_int8_deepseek_v3.py
@@ -16,7 +16,7 @@ from sglang.test.test_utils import (
 )
 
 # DeepSeek-V3 INT8 quantization tests (channel and block INT8)
-register_cuda_ci(est_time=160, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=160, stage="base-b", runner_config="1-gpu-large")
 
 
 class TestDeepseekV3MTPChannelInt8(CustomTestCase):
diff --git a/test/registered/model_loading/test_external_models.py b/test/registered/model_loading/test_external_models.py
index a03bffd33..59bef0663 100644
--- a/test/registered/model_loading/test_external_models.py
+++ b/test/registered/model_loading/test_external_models.py
@@ -5,7 +5,7 @@ from sglang.srt.environ import envs
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=29, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=29, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=45, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/model_loading/test_utils_update_weights.py b/test/registered/model_loading/test_utils_update_weights.py
index ff9bd3a54..89bca6240 100644
--- a/test/registered/model_loading/test_utils_update_weights.py
+++ b/test/registered/model_loading/test_utils_update_weights.py
@@ -12,7 +12,7 @@ from sglang.srt.weight_sync.utils import update_weights
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import DEFAULT_SMALL_MODEL_NAME_FOR_TEST
 
-register_cuda_ci(est_time=32, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=32, stage="base-b", runner_config="1-gpu-large")
 
 
 class AsyncEngine(Engine):
diff --git a/test/registered/models/test_dummy_grok_models.py b/test/registered/models/test_dummy_grok_models.py
index e251db8db..e98413e08 100644
--- a/test/registered/models/test_dummy_grok_models.py
+++ b/test/registered/models/test_dummy_grok_models.py
@@ -5,7 +5,7 @@ from sglang.test.test_utils import CustomTestCase, is_in_ci, run_bench_one_batch
 
 register_cuda_ci(
     est_time=120,
-    stage="stage-b",
+    stage="base-b",
     runner_config="2-gpu-large",
     disabled="Temporarily disabled",
 )
diff --git a/test/registered/models/test_gemma4_fp8_per_expert_loading.py b/test/registered/models/test_gemma4_fp8_per_expert_loading.py
index 114f29c9b..468b8baae 100644
--- a/test/registered/models/test_gemma4_fp8_per_expert_loading.py
+++ b/test/registered/models/test_gemma4_fp8_per_expert_loading.py
@@ -25,7 +25,7 @@ from sglang.test.test_utils import (
 
 # Compressed-tensors per-expert FP8 MoE checkpoint that exercises the
 # loader path (gated repo + ~27 GB download + 4 GPUs at TP=4).
-register_cuda_ci(est_time=120, stage="stage-c", runner_config="4-gpu-h100")
+register_cuda_ci(est_time=120, stage="base-c", runner_config="4-gpu-h100")
 
 
 @unittest.skipIf(get_device_sm() < 90, "Test requires CUDA SM 90 or higher")
diff --git a/test/registered/models/test_kimi_linear_models.py b/test/registered/models/test_kimi_linear_models.py
index bc86d5e42..1ce6d7c45 100644
--- a/test/registered/models/test_kimi_linear_models.py
+++ b/test/registered/models/test_kimi_linear_models.py
@@ -11,7 +11,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=178, stage="stage-b", runner_config="2-gpu-large")
+register_cuda_ci(est_time=178, stage="base-b", runner_config="2-gpu-large")
 
 
 class TestKimiLinear(CustomTestCase):
diff --git a/test/registered/models/test_ministral3_models.py b/test/registered/models/test_ministral3_models.py
index d9a174c2f..f99a2bc26 100644
--- a/test/registered/models/test_ministral3_models.py
+++ b/test/registered/models/test_ministral3_models.py
@@ -8,7 +8,7 @@ from sglang.test.server_fixtures.mmmu_fixture import MMMUServerBase
 
 register_cuda_ci(
     est_time=200,
-    stage="stage-b",
+    stage="base-b",
     runner_config="1-gpu-small",
     disabled="Temporarily disabled",
 )
diff --git a/test/registered/models/test_nvidia_nemotron_3_nano.py b/test/registered/models/test_nvidia_nemotron_3_nano.py
index 597eb5101..87e2f708f 100644
--- a/test/registered/models/test_nvidia_nemotron_3_nano.py
+++ b/test/registered/models/test_nvidia_nemotron_3_nano.py
@@ -6,7 +6,7 @@ from sglang.test.server_fixtures.default_fixture import DefaultServerBase
 
 register_cuda_ci(
     est_time=190,
-    stage="stage-b",
+    stage="base-b",
     runner_config="2-gpu-large",
 )
 
diff --git a/test/registered/models/test_transformers_backend_eval.py b/test/registered/models/test_transformers_backend_eval.py
index 2a1e96085..47a348848 100644
--- a/test/registered/models/test_transformers_backend_eval.py
+++ b/test/registered/models/test_transformers_backend_eval.py
@@ -7,7 +7,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.few_shot_gsm8k import run_eval
 from sglang.test.server_fixtures.default_fixture import DefaultServerBase
 
-register_cuda_ci(est_time=48, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=48, stage="base-b", runner_config="1-gpu-small")
 
 
 class TestTransformersBackendEval(DefaultServerBase):
diff --git a/test/registered/models/test_transformers_models.py b/test/registered/models/test_transformers_models.py
index 2b8095ff8..01e6a24d9 100644
--- a/test/registered/models/test_transformers_models.py
+++ b/test/registered/models/test_transformers_models.py
@@ -21,7 +21,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=177, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=177, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=320, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/moe/test_fused_moe.py b/test/registered/moe/test_fused_moe.py
index 75249bc7d..6a43bd7a2 100644
--- a/test/registered/moe/test_fused_moe.py
+++ b/test/registered/moe/test_fused_moe.py
@@ -13,7 +13,7 @@ from sglang.srt.utils import get_device, get_device_capability, is_hip
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase, empty_gpu_cache
 
-register_cuda_ci(est_time=87, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=87, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=30, suite="stage-b-test-1-gpu-small-amd")
 
 _is_hip = is_hip()
diff --git a/test/registered/moe/test_glm4_moe_models.py b/test/registered/moe/test_glm4_moe_models.py
index f4faffe3b..3553605cd 100644
--- a/test/registered/moe/test_glm4_moe_models.py
+++ b/test/registered/moe/test_glm4_moe_models.py
@@ -11,7 +11,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=171, stage="stage-b", runner_config="2-gpu-large")
+register_cuda_ci(est_time=171, stage="base-b", runner_config="2-gpu-large")
 
 
 class TestGLM4MoE(CustomTestCase):
diff --git a/test/registered/moe/test_moe_ep.py b/test/registered/moe/test_moe_ep.py
index cd1120603..6334a5d2b 100644
--- a/test/registered/moe/test_moe_ep.py
+++ b/test/registered/moe/test_moe_ep.py
@@ -14,7 +14,7 @@ from sglang.test.test_utils import (
 
 # Per-commit: TP=2 EP=2 baseline.
 # DeepGEMM/FP8 variant moved to test_moe_ep_nightly.py.
-register_cuda_ci(est_time=279, stage="stage-b", runner_config="2-gpu-large")
+register_cuda_ci(est_time=279, stage="base-b", runner_config="2-gpu-large")
 
 
 class TestEp(CustomTestCase):
diff --git a/test/registered/moe/test_torch_compile_moe.py b/test/registered/moe/test_torch_compile_moe.py
index 495ec9a5d..9e0816555 100644
--- a/test/registered/moe/test_torch_compile_moe.py
+++ b/test/registered/moe/test_torch_compile_moe.py
@@ -16,7 +16,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=130, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=130, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=1400, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/moe/test_triton_fused_moe.py b/test/registered/moe/test_triton_fused_moe.py
index da0a5195f..5edd08b1f 100644
--- a/test/registered/moe/test_triton_fused_moe.py
+++ b/test/registered/moe/test_triton_fused_moe.py
@@ -12,7 +12,7 @@ from sglang.srt.server_args import ServerArgs, set_global_server_args_for_schedu
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=13, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=13, stage="base-b", runner_config="1-gpu-large")
 
 
 class TestFusedMOE(CustomTestCase):
diff --git a/test/registered/moe/test_triton_moe_channel_fp8_kernel.py b/test/registered/moe/test_triton_moe_channel_fp8_kernel.py
index 2bbefdddb..f2bf09b20 100644
--- a/test/registered/moe/test_triton_moe_channel_fp8_kernel.py
+++ b/test/registered/moe/test_triton_moe_channel_fp8_kernel.py
@@ -11,7 +11,7 @@ from sglang.srt.server_args import ServerArgs, set_global_server_args_for_schedu
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=17, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=17, stage="base-b", runner_config="1-gpu-large")
 
 
 def native_w8a8_per_token_matmul(A, B, As, Bs, output_dtype=torch.float16):
diff --git a/test/registered/observability/test_metrics.py b/test/registered/observability/test_metrics.py
index d6db74295..7e812b881 100644
--- a/test/registered/observability/test_metrics.py
+++ b/test/registered/observability/test_metrics.py
@@ -20,7 +20,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=74, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=74, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=32, suite="stage-b-test-1-gpu-small-amd")
 
 _MODEL_NAME = "Qwen/Qwen3-0.6B"
diff --git a/test/registered/observability/test_priority_metrics.py b/test/registered/observability/test_priority_metrics.py
index fe778293e..0178d3006 100644
--- a/test/registered/observability/test_priority_metrics.py
+++ b/test/registered/observability/test_priority_metrics.py
@@ -18,7 +18,7 @@ from sglang.test.test_utils import (
 
 register_cuda_ci(
     est_time=60,
-    stage="stage-b",
+    stage="base-b",
     runner_config="1-gpu-small",
 )
 register_amd_ci(est_time=60, suite="stage-b-test-1-gpu-small-amd")
diff --git a/test/registered/observability/test_tracing_disaggregation.py b/test/registered/observability/test_tracing_disaggregation.py
index f16b23091..f38d513f3 100644
--- a/test/registered/observability/test_tracing_disaggregation.py
+++ b/test/registered/observability/test_tracing_disaggregation.py
@@ -33,7 +33,7 @@ from sglang.utils import wait_for_http_ready
 logger = logging.getLogger(__name__)
 
 # CI registration - PD disaggregation requires 2 GPUs
-register_cuda_ci(est_time=65, stage="stage-b", runner_config="2-gpu-large")
+register_cuda_ci(est_time=65, stage="base-b", runner_config="2-gpu-large")
 
 
 class TestTraceDisaggregation(CustomTestCase):
diff --git a/test/registered/openai_server/basic/test_anthropic_server.py b/test/registered/openai_server/basic/test_anthropic_server.py
index a77cd982c..6f9859072 100644
--- a/test/registered/openai_server/basic/test_anthropic_server.py
+++ b/test/registered/openai_server/basic/test_anthropic_server.py
@@ -31,7 +31,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=40, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=40, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=140, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/openai_server/basic/test_http2_server.py b/test/registered/openai_server/basic/test_http2_server.py
index 21b222684..e2c313114 100644
--- a/test/registered/openai_server/basic/test_http2_server.py
+++ b/test/registered/openai_server/basic/test_http2_server.py
@@ -27,7 +27,7 @@ try:
 except ImportError:
     _HAS_GRANIAN = False
 
-register_cuda_ci(est_time=52, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=52, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=52, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/openai_server/basic/test_openai_server.py b/test/registered/openai_server/basic/test_openai_server.py
index 682692022..7a9591764 100644
--- a/test/registered/openai_server/basic/test_openai_server.py
+++ b/test/registered/openai_server/basic/test_openai_server.py
@@ -28,7 +28,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=182, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=182, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=200, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/openai_server/basic/test_serving_transcription.py b/test/registered/openai_server/basic/test_serving_transcription.py
index 6add726f2..d4e5adb7f 100644
--- a/test/registered/openai_server/basic/test_serving_transcription.py
+++ b/test/registered/openai_server/basic/test_serving_transcription.py
@@ -21,7 +21,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=60, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=60, stage="base-b", runner_config="1-gpu-small")
 
 WHISPER_MODEL = "openai/whisper-large-v3"
 AUDIO_URL = "https://raw.githubusercontent.com/sgl-project/sgl-test-files/refs/heads/main/audios/Trump_WEF_2018_10s.mp3"
diff --git a/test/registered/openai_server/features/test_json_mode.py b/test/registered/openai_server/features/test_json_mode.py
index a24fda4c3..5cbb68a3d 100644
--- a/test/registered/openai_server/features/test_json_mode.py
+++ b/test/registered/openai_server/features/test_json_mode.py
@@ -14,7 +14,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=118, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=118, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=180, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/openai_server/features/test_openai_server_ebnf.py b/test/registered/openai_server/features/test_openai_server_ebnf.py
index 4188e841b..c81a1492a 100644
--- a/test/registered/openai_server/features/test_openai_server_ebnf.py
+++ b/test/registered/openai_server/features/test_openai_server_ebnf.py
@@ -13,7 +13,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=44, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=44, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=20, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/openai_server/features/test_openai_server_hidden_states.py b/test/registered/openai_server/features/test_openai_server_hidden_states.py
index 2eb5fd7b6..5146fc6cf 100644
--- a/test/registered/openai_server/features/test_openai_server_hidden_states.py
+++ b/test/registered/openai_server/features/test_openai_server_hidden_states.py
@@ -16,7 +16,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=222, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=222, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(
     est_time=186,
     suite="stage-b-test-1-gpu-small-amd",
diff --git a/test/registered/openai_server/function_call/test_anthropic_tool_use.py b/test/registered/openai_server/function_call/test_anthropic_tool_use.py
index 1a56b62c9..762c23ce3 100644
--- a/test/registered/openai_server/function_call/test_anthropic_tool_use.py
+++ b/test/registered/openai_server/function_call/test_anthropic_tool_use.py
@@ -25,7 +25,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=50, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=50, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=140, suite="stage-b-test-1-gpu-small-amd")
 
 # System message to guide Llama3.2 to produce proper tool call format
diff --git a/test/registered/openai_server/function_call/test_openai_function_calling.py b/test/registered/openai_server/function_call/test_openai_function_calling.py
index f4e248590..64e64b7f2 100644
--- a/test/registered/openai_server/function_call/test_openai_function_calling.py
+++ b/test/registered/openai_server/function_call/test_openai_function_calling.py
@@ -14,7 +14,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=100, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=100, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=73, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/openai_server/validation/test_large_max_new_tokens.py b/test/registered/openai_server/validation/test_large_max_new_tokens.py
index aa4446b76..8e88dd1f2 100644
--- a/test/registered/openai_server/validation/test_large_max_new_tokens.py
+++ b/test/registered/openai_server/validation/test_large_max_new_tokens.py
@@ -22,7 +22,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=58, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=58, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=41, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/openai_server/validation/test_matched_stop.py b/test/registered/openai_server/validation/test_matched_stop.py
index c9566191c..8284c8d07 100644
--- a/test/registered/openai_server/validation/test_matched_stop.py
+++ b/test/registered/openai_server/validation/test_matched_stop.py
@@ -10,7 +10,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=52, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=52, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=60, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/openai_server/validation/test_openai_server_ignore_eos.py b/test/registered/openai_server/validation/test_openai_server_ignore_eos.py
index a765f8950..ddd8ffcec 100644
--- a/test/registered/openai_server/validation/test_openai_server_ignore_eos.py
+++ b/test/registered/openai_server/validation/test_openai_server_ignore_eos.py
@@ -11,7 +11,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=44, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=44, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=47, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/openai_server/validation/test_request_length_validation.py b/test/registered/openai_server/validation/test_request_length_validation.py
index 1f5ae898d..1ee3ae9ad 100644
--- a/test/registered/openai_server/validation/test_request_length_validation.py
+++ b/test/registered/openai_server/validation/test_request_length_validation.py
@@ -12,7 +12,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=49, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=49, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=31, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/perf/test_vlm_perf_5090.py b/test/registered/perf/test_vlm_perf_5090.py
index 3c98f8568..02aa32af9 100644
--- a/test/registered/perf/test_vlm_perf_5090.py
+++ b/test/registered/perf/test_vlm_perf_5090.py
@@ -13,7 +13,7 @@ from sglang.test.test_utils import (
     write_github_step_summary,
 )
 
-register_cuda_ci(est_time=406, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=406, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=500, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/piecewise_cuda_graph/test_pcg_with_speculative_decoding.py b/test/registered/piecewise_cuda_graph/test_pcg_with_speculative_decoding.py
index 3bd1ef0da..91bab339e 100644
--- a/test/registered/piecewise_cuda_graph/test_pcg_with_speculative_decoding.py
+++ b/test/registered/piecewise_cuda_graph/test_pcg_with_speculative_decoding.py
@@ -11,7 +11,7 @@ import unittest
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.server_fixtures.pcg_spec_fixture import PCGSpecBase
 
-register_cuda_ci(est_time=531, stage="stage-b", runner_config="2-gpu-large")
+register_cuda_ci(est_time=531, stage="base-b", runner_config="2-gpu-large")
 
 
 class TestPCGWithEAGLE3(PCGSpecBase, unittest.TestCase):
diff --git a/test/registered/piecewise_cuda_graph/test_piecewise_cuda_graph_support_1_gpu.py b/test/registered/piecewise_cuda_graph/test_piecewise_cuda_graph_support_1_gpu.py
index abc44f285..11f4fbf1d 100644
--- a/test/registered/piecewise_cuda_graph/test_piecewise_cuda_graph_support_1_gpu.py
+++ b/test/registered/piecewise_cuda_graph/test_piecewise_cuda_graph_support_1_gpu.py
@@ -17,7 +17,7 @@ from sglang.test.test_utils import (
 )
 
 # CI Registration
-register_cuda_ci(est_time=180, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=180, stage="base-b", runner_config="1-gpu-large")
 
 
 class TestPiecewiseCudaGraphQwen25VL(CustomTestCase):
diff --git a/test/registered/prefill_only/test_embed_overrides.py b/test/registered/prefill_only/test_embed_overrides.py
index f1e67d262..0faeed8b8 100644
--- a/test/registered/prefill_only/test_embed_overrides.py
+++ b/test/registered/prefill_only/test_embed_overrides.py
@@ -24,7 +24,7 @@ from sglang.srt.server_args import MIS_DELIMITER_TOKEN_ID
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=9, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=9, suite="stage-b-test-1-gpu-small-amd")
 
 HIDDEN_DIM = 4
diff --git a/test/registered/prefill_only/test_embedding_models.py b/test/registered/prefill_only/test_embedding_models.py
index f52ca0410..a0a3d6c7b 100644
--- a/test/registered/prefill_only/test_embedding_models.py
+++ b/test/registered/prefill_only/test_embedding_models.py
@@ -35,7 +35,7 @@ register_amd_ci(
     suite="stage-b-test-1-gpu-small-amd",
     disabled="see https://github.com/sgl-project/sglang/issues/11127",
 )
-register_cuda_ci(est_time=136, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=136, stage="base-b", runner_config="1-gpu-small")
 
 MODEL_TO_CONFIG = {
     "Alibaba-NLP/gte-Qwen2-1.5B-instruct": (1, 1e-5),
diff --git a/test/registered/prefill_only/test_multi_item_scoring.py b/test/registered/prefill_only/test_multi_item_scoring.py
index 3a019affd..522bdc9b5 100644
--- a/test/registered/prefill_only/test_multi_item_scoring.py
+++ b/test/registered/prefill_only/test_multi_item_scoring.py
@@ -27,7 +27,7 @@ from sglang.test.test_utils import (
     CustomTestCase,
 )
 
-register_cuda_ci(est_time=211, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=211, stage="base-b", runner_config="1-gpu-small")
 
 TEST_MODEL_NAME = os.environ.get("TEST_MODEL_NAME", DEFAULT_SMALL_MODEL_NAME_FOR_TEST)
 TEST_CLASSIFICATION_BASE_MODEL = os.environ.get(
diff --git a/test/registered/prefill_only/test_openai_embedding.py b/test/registered/prefill_only/test_openai_embedding.py
index 8455358e0..0f6435be5 100644
--- a/test/registered/prefill_only/test_openai_embedding.py
+++ b/test/registered/prefill_only/test_openai_embedding.py
@@ -13,7 +13,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=91, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=91, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=141, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/prefill_only/test_pooled_hidden_states.py b/test/registered/prefill_only/test_pooled_hidden_states.py
index e112c13b1..52e51183e 100644
--- a/test/registered/prefill_only/test_pooled_hidden_states.py
+++ b/test/registered/prefill_only/test_pooled_hidden_states.py
@@ -27,7 +27,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=100, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=100, stage="base-b", runner_config="1-gpu-small")
 
 _SEQCLS_MODEL = "Qwen/Qwen3-0.6B"
 _CAUSAL_LM_MODEL = DEFAULT_SMALL_MODEL_NAME_FOR_TEST
diff --git a/test/registered/prefill_only/test_reward_models.py b/test/registered/prefill_only/test_reward_models.py
index 8647dba6a..b839a434e 100644
--- a/test/registered/prefill_only/test_reward_models.py
+++ b/test/registered/prefill_only/test_reward_models.py
@@ -24,7 +24,7 @@ from sglang.test.test_utils import CustomTestCase
 # ==============================================================================
 
 
-register_cuda_ci(est_time=166, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=166, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=132, suite="stage-b-test-1-gpu-small-amd-nondeterministic")
 
 MODELS = [
diff --git a/test/registered/prefill_only/test_score_api.py b/test/registered/prefill_only/test_score_api.py
index 7e6059588..0b5f07403 100644
--- a/test/registered/prefill_only/test_score_api.py
+++ b/test/registered/prefill_only/test_score_api.py
@@ -27,7 +27,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=71, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=71, stage="base-b", runner_config="1-gpu-small")
 
 _MODEL = os.environ.get("TEST_MODEL_NAME", DEFAULT_SMALL_MODEL_NAME_FOR_TEST)
 
diff --git a/test/registered/prefill_only/test_score_engine.py b/test/registered/prefill_only/test_score_engine.py
index ff0550f2f..959ec28c8 100644
--- a/test/registered/prefill_only/test_score_engine.py
+++ b/test/registered/prefill_only/test_score_engine.py
@@ -26,7 +26,7 @@ from sglang.srt.entrypoints.engine import Engine
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import DEFAULT_SMALL_MODEL_NAME_FOR_TEST, CustomTestCase
 
-register_cuda_ci(est_time=85, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=85, stage="base-b", runner_config="1-gpu-small")
 
 _CAUSAL_LM_MODEL = os.environ.get("TEST_MODEL_NAME", DEFAULT_SMALL_MODEL_NAME_FOR_TEST)
 _SEQCLS_MODEL = os.environ.get("TEST_CLASSIFICATION_BASE_MODEL", "Qwen/Qwen3-0.6B")
diff --git a/test/registered/prefill_only/test_serving_rerank.py b/test/registered/prefill_only/test_serving_rerank.py
index 216278a99..0555d5f6a 100644
--- a/test/registered/prefill_only/test_serving_rerank.py
+++ b/test/registered/prefill_only/test_serving_rerank.py
@@ -7,7 +7,7 @@ from sglang.srt.managers.tokenizer_manager_score_mixin import ScoreResult
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 # Keep consistent with other openai_server/basic unit tests.
-register_cuda_ci(est_time=9, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=10, suite="stage-b-test-1-gpu-small-amd")
 
 try:
diff --git a/test/registered/profiling/test_profile_v2.py b/test/registered/profiling/test_profile_v2.py
index ac9d6b2e0..17b194f2d 100644
--- a/test/registered/profiling/test_profile_v2.py
+++ b/test/registered/profiling/test_profile_v2.py
@@ -19,7 +19,7 @@ from sglang.test.test_utils import (
 
 register_cuda_ci(
     est_time=120,
-    stage="stage-b",
+    stage="base-b",
     runner_config="1-gpu-small",
     disabled="Temporarily disabled",
 )
diff --git a/test/registered/profiling/test_start_profile.py b/test/registered/profiling/test_start_profile.py
index 34630c922..df7fd579a 100644
--- a/test/registered/profiling/test_start_profile.py
+++ b/test/registered/profiling/test_start_profile.py
@@ -29,7 +29,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=42, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=42, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=60, suite="stage-b-test-1-gpu-small-amd")
 
 OUTPUT_DIR = "./profiler_dir"
diff --git a/test/registered/quant/test_awq.py b/test/registered/quant/test_awq.py
index fbdb69e35..abaadd7ba 100644
--- a/test/registered/quant/test_awq.py
+++ b/test/registered/quant/test_awq.py
@@ -13,7 +13,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=160, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=160, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=200, suite="stage-b-test-1-gpu-large-amd")
 
 
diff --git a/test/registered/quant/test_block_int8.py b/test/registered/quant/test_block_int8.py
index 2c1d77def..dc456b2f4 100644
--- a/test/registered/quant/test_block_int8.py
+++ b/test/registered/quant/test_block_int8.py
@@ -10,7 +10,7 @@ from sglang.srt.server_args import ServerArgs, set_global_server_args_for_schedu
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=44, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=44, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=22, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/quant/test_deepseek_v32_fp4_mtp_4gpu.py b/test/registered/quant/test_deepseek_v32_fp4_mtp_4gpu.py
index 6ded6338b..2ef4633e7 100644
--- a/test/registered/quant/test_deepseek_v32_fp4_mtp_4gpu.py
+++ b/test/registered/quant/test_deepseek_v32_fp4_mtp_4gpu.py
@@ -17,7 +17,7 @@ from sglang.test.test_utils import (
 
 register_cuda_ci(
     est_time=690,
-    stage="stage-c",
+    stage="base-c",
     runner_config="4-gpu-b200",
 )
 
diff --git a/test/registered/quant/test_deepseek_v3_fp4_4gpu.py b/test/registered/quant/test_deepseek_v3_fp4_4gpu.py
index 5eadc8287..927636018 100644
--- a/test/registered/quant/test_deepseek_v3_fp4_4gpu.py
+++ b/test/registered/quant/test_deepseek_v3_fp4_4gpu.py
@@ -15,7 +15,7 @@ from sglang.test.test_utils import (
 # Per-commit: SymmetricMemory variant only.
 # - TestDeepseekV3FP4 (TRTLLM) archived to test/manual/quant/test_deepseek_v3_fp4_4gpu_trtllm.py
 # - TestDeepseekV3FP4CutlassMoE moved to test_deepseek_v3_fp4_4gpu_extra.py
-register_cuda_ci(est_time=960, stage="stage-c", runner_config="4-gpu-b200")
+register_cuda_ci(est_time=960, stage="base-c", runner_config="4-gpu-b200")
 
 FULL_DEEPSEEK_V3_FP4_MODEL_PATH = "nvidia/DeepSeek-V3-0324-FP4"
 SERVER_LAUNCH_TIMEOUT = 1200
diff --git a/test/registered/quant/test_fp8_kernel.py b/test/registered/quant/test_fp8_kernel.py
index a1cf1fa84..9379bc257 100644
--- a/test/registered/quant/test_fp8_kernel.py
+++ b/test/registered/quant/test_fp8_kernel.py
@@ -9,7 +9,7 @@ from sglang.srt.layers.quantization.fp8_kernel import (
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=10, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-large")
 
 from sglang.srt.utils import get_device, is_cuda, is_xpu
 
diff --git a/test/registered/quant/test_fp8_utils.py b/test/registered/quant/test_fp8_utils.py
index 277051c04..34dbc9ccd 100644
--- a/test/registered/quant/test_fp8_utils.py
+++ b/test/registered/quant/test_fp8_utils.py
@@ -10,7 +10,7 @@ from sglang.srt.layers.quantization.fp8_utils import (
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=9, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-large")
 
 
 class TestInverseTransformScaleUe8m0(CustomTestCase):
diff --git a/test/registered/quant/test_gguf.py b/test/registered/quant/test_gguf.py
index 50ed3bdab..083d814be 100644
--- a/test/registered/quant/test_gguf.py
+++ b/test/registered/quant/test_gguf.py
@@ -6,7 +6,7 @@ import sglang as sgl
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=76, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=76, stage="base-b", runner_config="1-gpu-small")
 
 
 class TestGGUF(CustomTestCase):
diff --git a/test/registered/quant/test_int8_kernel.py b/test/registered/quant/test_int8_kernel.py
index b3145f8f5..02cb34af1 100644
--- a/test/registered/quant/test_int8_kernel.py
+++ b/test/registered/quant/test_int8_kernel.py
@@ -11,7 +11,7 @@ from sglang.srt.server_args import ServerArgs, set_global_server_args_for_schedu
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=15, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=15, stage="base-b", runner_config="1-gpu-small")
 
 
 def native_w8a8_per_token_matmul(A, B, As, Bs, output_dtype=torch.float16):
diff --git a/test/registered/quant/test_is_layer_skipped.py b/test/registered/quant/test_is_layer_skipped.py
index 89e2c15ed..c311fc576 100644
--- a/test/registered/quant/test_is_layer_skipped.py
+++ b/test/registered/quant/test_is_layer_skipped.py
@@ -4,7 +4,7 @@ from sglang.srt.layers.quantization.utils import is_layer_skipped
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=5, suite="base-a-test-cpu")
 
 
 # Qwen3-Next FP8 actually publishes the equivalent of this in
diff --git a/test/registered/quant/test_marlin_moe.py b/test/registered/quant/test_marlin_moe.py
index f7aaeac0b..86d2ca837 100644
--- a/test/registered/quant/test_marlin_moe.py
+++ b/test/registered/quant/test_marlin_moe.py
@@ -12,7 +12,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_marlin_utils import awq_marlin_quantize, marlin_quantize
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=108, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=108, stage="base-b", runner_config="1-gpu-small")
 
 set_global_server_args_for_scheduler(object.__new__(ServerArgs))
 
diff --git a/test/registered/quant/test_modelopt_fp8.py b/test/registered/quant/test_modelopt_fp8.py
index 79e1bd9b2..d753b559f 100644
--- a/test/registered/quant/test_modelopt_fp8.py
+++ b/test/registered/quant/test_modelopt_fp8.py
@@ -12,7 +12,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=62, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=62, stage="base-b", runner_config="1-gpu-large")
 
 
 class TestModeloptFP8(CustomTestCase):
diff --git a/test/registered/quant/test_nvfp4_gemm.py b/test/registered/quant/test_nvfp4_gemm.py
index c8b39167c..06d4cbc52 100644
--- a/test/registered/quant/test_nvfp4_gemm.py
+++ b/test/registered/quant/test_nvfp4_gemm.py
@@ -12,7 +12,7 @@ from sglang.test.test_utils import (
     try_cached_model,
 )
 
-register_cuda_ci(est_time=350, stage="stage-c", runner_config="4-gpu-b200")
+register_cuda_ci(est_time=350, stage="base-c", runner_config="4-gpu-b200")
 
 MODEL_PATH = "nvidia/Llama-3.1-8B-Instruct-NVFP4"
 
diff --git a/test/registered/quant/test_nvfp4_gemm_sm120.py b/test/registered/quant/test_nvfp4_gemm_sm120.py
index 4dc6fa348..61dd5ea88 100644
--- a/test/registered/quant/test_nvfp4_gemm_sm120.py
+++ b/test/registered/quant/test_nvfp4_gemm_sm120.py
@@ -12,7 +12,7 @@ from sglang.test.test_utils import (
     try_cached_model,
 )
 
-register_cuda_ci(est_time=109, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=109, stage="base-b", runner_config="1-gpu-small")
 
 MODEL_PATH = "nvidia/Llama-3.1-8B-Instruct-NVFP4"
 
diff --git a/test/registered/quant/test_quant_config_parsing.py b/test/registered/quant/test_quant_config_parsing.py
index 33ebda5b8..ee30eebd0 100644
--- a/test/registered/quant/test_quant_config_parsing.py
+++ b/test/registered/quant/test_quant_config_parsing.py
@@ -5,7 +5,7 @@ from sglang.srt.configs.model_config import ModelConfig
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=15, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=15, suite="base-a-test-cpu")
 
 
 class TestQuantLogString(CustomTestCase):
diff --git a/test/registered/quant/test_triton_scaled_mm.py b/test/registered/quant/test_triton_scaled_mm.py
index da353db16..3d5dc8022 100644
--- a/test/registered/quant/test_triton_scaled_mm.py
+++ b/test/registered/quant/test_triton_scaled_mm.py
@@ -9,7 +9,7 @@ from sglang.srt.utils.common import get_device
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=11, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=12, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/radix_cache/test_radix_attention.py b/test/registered/radix_cache/test_radix_attention.py
index b88dccfc4..8814f6a27 100644
--- a/test/registered/radix_cache/test_radix_attention.py
+++ b/test/registered/radix_cache/test_radix_attention.py
@@ -14,7 +14,7 @@ from sglang.test.test_utils import (
 )
 
 # RadixAttention server integration tests
-register_cuda_ci(est_time=100, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=100, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=100, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/radix_cache/test_radix_cache_hit.py b/test/registered/radix_cache/test_radix_cache_hit.py
index de465f2dc..c052dbfdf 100644
--- a/test/registered/radix_cache/test_radix_cache_hit.py
+++ b/test/registered/radix_cache/test_radix_cache_hit.py
@@ -11,7 +11,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=55, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=55, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=55, suite="stage-b-test-1-gpu-small-amd")
 
 MODEL = DEFAULT_SMALL_MODEL_NAME_FOR_TEST
diff --git a/test/registered/radix_cache/test_swa_radix_cache_kl.py b/test/registered/radix_cache/test_swa_radix_cache_kl.py
index d2d66fdf5..795ad9d9e 100644
--- a/test/registered/radix_cache/test_swa_radix_cache_kl.py
+++ b/test/registered/radix_cache/test_swa_radix_cache_kl.py
@@ -6,7 +6,7 @@ from sglang.test.server_fixtures.default_fixture import DefaultServerBase
 
 MODEL = "openai/gpt-oss-20b"
 
-register_cuda_ci(est_time=151, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=151, stage="base-b", runner_config="1-gpu-large")
 
 
 class TestSWARadixCacheKL(KLDivergenceMixin, DefaultServerBase):
diff --git a/test/registered/radix_cache/test_unified_radix_cache_kl.py b/test/registered/radix_cache/test_unified_radix_cache_kl.py
index e6d38c6a2..2d8ab360f 100644
--- a/test/registered/radix_cache/test_unified_radix_cache_kl.py
+++ b/test/registered/radix_cache/test_unified_radix_cache_kl.py
@@ -35,7 +35,7 @@ MAMBA_TRACK_INTERVAL = 128
 SWA_MODEL = "openai/gpt-oss-20b"
 FULL_MODEL = "Qwen/Qwen3-32B"
 
-register_cuda_ci(est_time=760, stage="stage-c", runner_config="4-gpu-h100")
+register_cuda_ci(est_time=760, stage="base-c", runner_config="4-gpu-h100")
 
 
 class UnifiedRadixTreeTestMixin:
diff --git a/test/registered/reasoning/test_reasoning.py b/test/registered/reasoning/test_reasoning.py
index 2bd16c387..ac570c635 100644
--- a/test/registered/reasoning/test_reasoning.py
+++ b/test/registered/reasoning/test_reasoning.py
@@ -17,7 +17,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=129, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=129, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=200, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/rl/test_fp32_lm_head.py b/test/registered/rl/test_fp32_lm_head.py
index b6aeb4ca9..dac52502c 100644
--- a/test/registered/rl/test_fp32_lm_head.py
+++ b/test/registered/rl/test_fp32_lm_head.py
@@ -15,7 +15,7 @@ from sglang.srt.server_args import (
 from sglang.srt.utils import get_device
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=9, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=15, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/rl/test_patch_torch.py b/test/registered/rl/test_patch_torch.py
index 6844a856e..fd5dc9412 100644
--- a/test/registered/rl/test_patch_torch.py
+++ b/test/registered/rl/test_patch_torch.py
@@ -9,7 +9,7 @@ import torch.multiprocessing as mp
 from sglang.srt.utils.patch_torch import monkey_patch_torch_reductions
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=15, stage="stage-b", runner_config="2-gpu-large")
+register_cuda_ci(est_time=15, stage="base-b", runner_config="2-gpu-large")
 
 
 class TestReleaseMemoryOccupation(unittest.TestCase):
diff --git a/test/registered/rl/test_pause_generation_tensor_consistency.py b/test/registered/rl/test_pause_generation_tensor_consistency.py
index 64c172221..5c1f54032 100644
--- a/test/registered/rl/test_pause_generation_tensor_consistency.py
+++ b/test/registered/rl/test_pause_generation_tensor_consistency.py
@@ -9,7 +9,7 @@ import torch
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 
 # ---------------------------------------------------------------------------
diff --git a/test/registered/rl/test_release_memory_occupation.py b/test/registered/rl/test_release_memory_occupation.py
index 4e2f024a8..8628114b1 100644
--- a/test/registered/rl/test_release_memory_occupation.py
+++ b/test/registered/rl/test_release_memory_occupation.py
@@ -55,7 +55,7 @@ from sglang.test.test_utils import (
 
 register_cuda_ci(
     est_time=200,
-    stage="stage-c",
+    stage="base-c",
     runner_config="4-gpu-h100",
     disabled="Temporarily disabled - needs investigation",
 )
diff --git a/test/registered/rl/test_update_weights_from_disk.py b/test/registered/rl/test_update_weights_from_disk.py
index 415aee265..ed7255916 100644
--- a/test/registered/rl/test_update_weights_from_disk.py
+++ b/test/registered/rl/test_update_weights_from_disk.py
@@ -22,7 +22,7 @@ register_amd_ci(
     est_time=210, suite="stage-b-test-1-gpu-small-amd", disabled="see #14021"
 )
 register_cuda_ci(
-    est_time=210, stage="stage-b", runner_config="1-gpu-large", disabled="see #14021"
+    est_time=210, stage="base-b", runner_config="1-gpu-large", disabled="see #14021"
 )
 
 
diff --git a/test/registered/sampling/test_original_logprobs.py b/test/registered/sampling/test_original_logprobs.py
index cefc40256..c0e4263d1 100644
--- a/test/registered/sampling/test_original_logprobs.py
+++ b/test/registered/sampling/test_original_logprobs.py
@@ -25,7 +25,7 @@ import sglang as sgl
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import DEFAULT_SMALL_MODEL_NAME_FOR_TEST
 
-register_cuda_ci(est_time=45, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=45, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=60, suite="stage-b-test-1-gpu-small-amd")
 
 # ------------------------- Configurable via env ------------------------- #
diff --git a/test/registered/sampling/test_penalty.py b/test/registered/sampling/test_penalty.py
index 39e42a552..63f7bc70a 100644
--- a/test/registered/sampling/test_penalty.py
+++ b/test/registered/sampling/test_penalty.py
@@ -16,7 +16,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=53, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=53, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=82, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/sampling/test_pytorch_sampling_backend.py b/test/registered/sampling/test_pytorch_sampling_backend.py
index 73ff86d05..4322df298 100644
--- a/test/registered/sampling/test_pytorch_sampling_backend.py
+++ b/test/registered/sampling/test_pytorch_sampling_backend.py
@@ -15,7 +15,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=80, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=80, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=66, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/scheduler/test_abort_with_metrics.py b/test/registered/scheduler/test_abort_with_metrics.py
index 4f894a31f..17d875b4c 100644
--- a/test/registered/scheduler/test_abort_with_metrics.py
+++ b/test/registered/scheduler/test_abort_with_metrics.py
@@ -18,7 +18,7 @@ from sglang.srt.utils.http_middleware_patch import _PureASGIDispatch
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 _HTTP_SCOPE = {
     "type": "http",
diff --git a/test/registered/scheduler/test_mixed_chunked_prefill.py b/test/registered/scheduler/test_mixed_chunked_prefill.py
index fcfd73ce9..15c448d92 100644
--- a/test/registered/scheduler/test_mixed_chunked_prefill.py
+++ b/test/registered/scheduler/test_mixed_chunked_prefill.py
@@ -12,7 +12,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=167, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=167, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=180, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/scheduler/test_prefill_delayer.py b/test/registered/scheduler/test_prefill_delayer.py
index 24ea98015..c4ba97e0f 100644
--- a/test/registered/scheduler/test_prefill_delayer.py
+++ b/test/registered/scheduler/test_prefill_delayer.py
@@ -28,7 +28,7 @@ from sglang.test.test_utils import (
 
 register_cuda_ci(
     est_time=300,
-    stage="stage-c",
+    stage="base-c",
     runner_config="8-gpu-h200",
     disabled="Temporarily disabled",
 )
diff --git a/test/registered/scheduler/test_retract_decode.py b/test/registered/scheduler/test_retract_decode.py
index 4db2c790b..31b9f6715 100644
--- a/test/registered/scheduler/test_retract_decode.py
+++ b/test/registered/scheduler/test_retract_decode.py
@@ -17,7 +17,7 @@ from sglang.test.test_utils import (
 )
 from sglang.utils import is_in_ci
 
-register_cuda_ci(est_time=353, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=353, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=600, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/scheduler/test_scheduler_control.py b/test/registered/scheduler/test_scheduler_control.py
index 0ef3bf66e..aa002bb54 100644
--- a/test/registered/scheduler/test_scheduler_control.py
+++ b/test/registered/scheduler/test_scheduler_control.py
@@ -20,7 +20,7 @@ from sglang.test.test_utils import (
     run_and_check_memory_leak,
 )
 
-register_cuda_ci(est_time=367, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=367, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=300, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/sessions/test_streaming_session.py b/test/registered/sessions/test_streaming_session.py
index 9fce4c94b..8485b1c7c 100644
--- a/test/registered/sessions/test_streaming_session.py
+++ b/test/registered/sessions/test_streaming_session.py
@@ -23,7 +23,7 @@ from sglang.test.test_utils import (
     DEFAULT_TARGET_MODEL_EAGLE3,
 )
 
-register_cuda_ci(est_time=691, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=691, stage="base-b", runner_config="1-gpu-large")
 
 
 class TestStreamingSession(StreamingSessionServerBase, StreamingSessionKitMixin):
diff --git a/test/registered/sessions/test_streaming_session_swa.py b/test/registered/sessions/test_streaming_session_swa.py
index b9b17d4d6..889a12cf2 100644
--- a/test/registered/sessions/test_streaming_session_swa.py
+++ b/test/registered/sessions/test_streaming_session_swa.py
@@ -12,7 +12,7 @@ from sglang.test.server_fixtures.streaming_session_fixture import (
     StreamingSessionServerBase,
 )
 
-register_cuda_ci(est_time=519, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=519, stage="base-b", runner_config="1-gpu-large")
 
 
 SWA_MODEL = "openai/gpt-oss-20b"
diff --git a/test/registered/spec/dflash/test_dflash.py b/test/registered/spec/dflash/test_dflash.py
index ef6ad8f84..d139cbf54 100644
--- a/test/registered/spec/dflash/test_dflash.py
+++ b/test/registered/spec/dflash/test_dflash.py
@@ -18,7 +18,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=302, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=302, stage="base-b", runner_config="1-gpu-small")
 
 
 class TestDFlashServerBase(CustomTestCase, MatchedStopMixin, GSM8KMixin):
diff --git a/test/registered/spec/eagle/test_adaptive_speculative.py b/test/registered/spec/eagle/test_adaptive_speculative.py
index 5a1972201..97ac8b900 100644
--- a/test/registered/spec/eagle/test_adaptive_speculative.py
+++ b/test/registered/spec/eagle/test_adaptive_speculative.py
@@ -18,7 +18,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=76, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=76, stage="base-b", runner_config="1-gpu-large")
 
 HIGH_ACCEPT_PROMPT = (
     "Output exactly 128 new lines. "
diff --git a/test/registered/spec/eagle/test_deepseek_v3_fp4_mtp_small.py b/test/registered/spec/eagle/test_deepseek_v3_fp4_mtp_small.py
index faa4ed476..99b3a2e88 100644
--- a/test/registered/spec/eagle/test_deepseek_v3_fp4_mtp_small.py
+++ b/test/registered/spec/eagle/test_deepseek_v3_fp4_mtp_small.py
@@ -16,7 +16,7 @@ from sglang.test.test_utils import (
     write_github_step_summary,
 )
 
-register_cuda_ci(est_time=340, stage="stage-b", runner_config="4-gpu-b200")
+register_cuda_ci(est_time=340, stage="base-b", runner_config="4-gpu-b200")
 
 FULL_DEEPSEEK_V3_FP4_MODEL_PATH = "nvidia/DeepSeek-V3-0324-FP4"
 SERVER_LAUNCH_TIMEOUT = 1200
diff --git a/test/registered/spec/eagle/test_eagle_constrained_decoding.py b/test/registered/spec/eagle/test_eagle_constrained_decoding.py
index 4087332c6..6478ce0b3 100644
--- a/test/registered/spec/eagle/test_eagle_constrained_decoding.py
+++ b/test/registered/spec/eagle/test_eagle_constrained_decoding.py
@@ -14,7 +14,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=116, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=116, stage="base-b", runner_config="1-gpu-large")
 
 
 class TestEagleConstrainedDecoding(
diff --git a/test/registered/spec/eagle/test_eagle_infer_b.py b/test/registered/spec/eagle/test_eagle_infer_b.py
index aa6ce7356..192b75c3f 100644
--- a/test/registered/spec/eagle/test_eagle_infer_b.py
+++ b/test/registered/spec/eagle/test_eagle_infer_b.py
@@ -22,7 +22,7 @@ from sglang.test.run_eval import run_eval
 from sglang.test.server_fixtures.eagle_fixture import EagleServerBase
 from sglang.test.test_utils import DEFAULT_TARGET_MODEL_EAGLE, run_logprob_check
 
-register_cuda_ci(est_time=847, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=847, stage="base-b", runner_config="1-gpu-large")
 
 
 class TestEAGLEServerBasic(EagleServerBase):
diff --git a/test/registered/spec/eagle/test_eagle_infer_beta.py b/test/registered/spec/eagle/test_eagle_infer_beta.py
index 75635ee63..92577ca0f 100644
--- a/test/registered/spec/eagle/test_eagle_infer_beta.py
+++ b/test/registered/spec/eagle/test_eagle_infer_beta.py
@@ -19,7 +19,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=369, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=369, stage="base-b", runner_config="1-gpu-small")
 
 
 class TestEagle3ServerBase(CustomTestCase, MatchedStopMixin):
diff --git a/test/registered/spec/eagle/test_eagle_infer_beta_dp_attention.py b/test/registered/spec/eagle/test_eagle_infer_beta_dp_attention.py
index 3f9906b9f..9e4bf5676 100644
--- a/test/registered/spec/eagle/test_eagle_infer_beta_dp_attention.py
+++ b/test/registered/spec/eagle/test_eagle_infer_beta_dp_attention.py
@@ -17,7 +17,7 @@ from sglang.test.test_utils import (
 )
 
 # EAGLE with DP attention on B200 (tp=2, dp=2, requires 4 B200 GPUs)
-register_cuda_ci(est_time=90, stage="stage-c", runner_config="4-gpu-b200")
+register_cuda_ci(est_time=90, stage="base-c", runner_config="4-gpu-b200")
 
 
 def test_gsm8k(base_url: str, model: str):
diff --git a/test/registered/spec/test_constrained_decoding_spec_reasoning.py b/test/registered/spec/test_constrained_decoding_spec_reasoning.py
index 4f1322ea4..60e695bc0 100644
--- a/test/registered/spec/test_constrained_decoding_spec_reasoning.py
+++ b/test/registered/spec/test_constrained_decoding_spec_reasoning.py
@@ -14,7 +14,7 @@ from sglang.test.test_utils import (
 )
 
 # Constrained decoding with EAGLE3 speculative reasoning (tp=2)
-register_cuda_ci(est_time=137, stage="stage-b", runner_config="2-gpu-large")
+register_cuda_ci(est_time=137, stage="base-b", runner_config="2-gpu-large")
 
 
 class ServerWithGrammar(CustomTestCase):
diff --git a/test/registered/spec/test_spec_ngram.py b/test/registered/spec/test_spec_ngram.py
index 65e001c55..02b21928a 100644
--- a/test/registered/spec/test_spec_ngram.py
+++ b/test/registered/spec/test_spec_ngram.py
@@ -7,7 +7,7 @@ from sglang.test.server_fixtures.ngram_fixture import NgramServerBase
 # Per-commit: Paged backend only.
 # - FA3 base test archived to test/manual/spec/test_spec_ngram_fa3.py
 # - Triton + Flashinfer moved to test_spec_ngram_extra.py
-register_cuda_ci(est_time=254, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=254, stage="base-b", runner_config="1-gpu-large")
 
 
 class TestNgramSpeculativeDecodingPaged(NgramServerBase, GSM8KMixin):
diff --git a/test/registered/spec/test_spec_standalone.py b/test/registered/spec/test_spec_standalone.py
index afd45017a..f86f6f2c5 100644
--- a/test/registered/spec/test_spec_standalone.py
+++ b/test/registered/spec/test_spec_standalone.py
@@ -6,7 +6,7 @@ from sglang.test.test_utils import CustomTestCase
 
 # V2 standalone speculative decoding tests (FA3, Triton, FlashInfer backends).
 # Non-V2 backends moved to test_spec_standalone_extra.py.
-register_cuda_ci(est_time=406, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=406, stage="base-b", runner_config="1-gpu-large")
 
 
 class TestStandaloneV2SpeculativeDecodingBase(StandaloneServerBase, CustomTestCase):
diff --git a/test/registered/spec/utils/test_build_eagle_tree.py b/test/registered/spec/utils/test_build_eagle_tree.py
index 809f2d1d2..9da57cfa5 100644
--- a/test/registered/spec/utils/test_build_eagle_tree.py
+++ b/test/registered/spec/utils/test_build_eagle_tree.py
@@ -9,7 +9,7 @@ from sglang.srt.speculative.eagle_utils import (
 from sglang.srt.utils import get_device
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=6, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=6, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=3, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/tokenizer/test_multi_tokenizer.py b/test/registered/tokenizer/test_multi_tokenizer.py
index df2bb3aa0..1454679d1 100644
--- a/test/registered/tokenizer/test_multi_tokenizer.py
+++ b/test/registered/tokenizer/test_multi_tokenizer.py
@@ -17,7 +17,7 @@ from sglang.test.test_utils import (
     write_github_step_summary,
 )
 
-register_cuda_ci(est_time=211, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=211, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=345, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/tokenizer/test_skip_tokenizer_init.py b/test/registered/tokenizer/test_skip_tokenizer_init.py
index a4275801a..72802bedc 100644
--- a/test/registered/tokenizer/test_skip_tokenizer_init.py
+++ b/test/registered/tokenizer/test_skip_tokenizer_init.py
@@ -23,7 +23,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=79, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=79, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=117, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/unit/README.md b/test/registered/unit/README.md
index a8e5a8174..711e729f3 100644
--- a/test/registered/unit/README.md
+++ b/test/registered/unit/README.md
@@ -14,8 +14,8 @@ Tests can use CPU or GPU — the key criterion is **no server process**.
 3. Register for CI at the **top of the file** (after imports, before test classes):
    ```python
    from sglang.test.ci.ci_register import register_cpu_ci
-   register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
-   # or: register_cuda_ci(est_time=10, stage="stage-b", runner_config="1-gpu-small")
+   register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+   # or: register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small")
    ```
    CUDA suites whose names follow `{stage}-test-{runner_config}` use the
    `stage=` + `runner_config=` form. Legacy `suite="..."` is kept for
@@ -44,7 +44,7 @@ Tests can use CPU or GPU — the key criterion is **no server process**.
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=5, suite="base-a-test-cpu")
 
 import unittest
 
@@ -81,7 +81,7 @@ maybe_stub_sgl_kernel()  # must precede any import that pulls in sgl_kernel
 from sglang.srt.managers.io_struct import FlushCacheReqInput
 from sglang.srt.managers.scheduler import Scheduler
 
-register_cpu_ci(est_time=2, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=2, suite="base-a-test-cpu")
 ```
 
 The same pattern (`sys.meta_path` finder) can be applied to other GPU-only packages.
diff --git a/test/registered/unit/auto_benchmark/test_dataset_tools.py b/test/registered/unit/auto_benchmark/test_dataset_tools.py
index 37e49224b..1f47b1ead 100644
--- a/test/registered/unit/auto_benchmark/test_dataset_tools.py
+++ b/test/registered/unit/auto_benchmark/test_dataset_tools.py
@@ -15,7 +15,7 @@ from sglang.auto_benchmark_lib import infer_backend, prepare_dataset
 from sglang.benchmark.datasets.autobench import sample_autobench_requests
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=6, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=6, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=6, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/unit/auto_benchmark/test_run_candidate.py b/test/registered/unit/auto_benchmark/test_run_candidate.py
index 427f83220..9b9813d5c 100644
--- a/test/registered/unit/auto_benchmark/test_run_candidate.py
+++ b/test/registered/unit/auto_benchmark/test_run_candidate.py
@@ -14,7 +14,7 @@ from auto_benchmark import AutoBenchmarkTestCase
 from sglang.auto_benchmark_lib import SearchDeadlineExceeded, run_candidate
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=6, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=6, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=6, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/unit/auto_benchmark/test_search_tools.py b/test/registered/unit/auto_benchmark/test_search_tools.py
index b142492b4..b65db2098 100644
--- a/test/registered/unit/auto_benchmark/test_search_tools.py
+++ b/test/registered/unit/auto_benchmark/test_search_tools.py
@@ -28,7 +28,7 @@ from sglang.auto_benchmark_lib import (
 )
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=6, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=6, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=6, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/unit/bench/test_mmmu_eval_utils.py b/test/registered/unit/bench/test_mmmu_eval_utils.py
index fe6a16ae3..d6de7fe04 100644
--- a/test/registered/unit/bench/test_mmmu_eval_utils.py
+++ b/test/registered/unit/bench/test_mmmu_eval_utils.py
@@ -15,7 +15,7 @@ except ModuleNotFoundError:
         pass
 
 
-register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=5, suite="base-a-test-cpu")
 
 
 def _load_mmmu_eval_utils():
diff --git a/test/registered/unit/configs/test_linear_attn_model_registry.py b/test/registered/unit/configs/test_linear_attn_model_registry.py
index 9eb6970ca..6cb570b52 100644
--- a/test/registered/unit/configs/test_linear_attn_model_registry.py
+++ b/test/registered/unit/configs/test_linear_attn_model_registry.py
@@ -13,7 +13,7 @@ from sglang.srt.configs.linear_attn_model_registry import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 
 # Dummy config classes for testing
diff --git a/test/registered/unit/configs/test_model_config_parser_registry.py b/test/registered/unit/configs/test_model_config_parser_registry.py
index 3c1329858..01e319989 100644
--- a/test/registered/unit/configs/test_model_config_parser_registry.py
+++ b/test/registered/unit/configs/test_model_config_parser_registry.py
@@ -13,7 +13,7 @@ from sglang.srt.configs.model_config_parser_registry import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=1, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=1, suite="base-a-test-cpu")
 
 
 class _FakeParser(ModelConfigParserBase):
diff --git a/test/registered/unit/constrained/test_base_grammar_backend.py b/test/registered/unit/constrained/test_base_grammar_backend.py
index 191fb4b3a..f9cb636ca 100644
--- a/test/registered/unit/constrained/test_base_grammar_backend.py
+++ b/test/registered/unit/constrained/test_base_grammar_backend.py
@@ -30,7 +30,7 @@ from sglang.srt.constrained.base_grammar_backend import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(2.0, "stage-a-test-cpu")
+register_cpu_ci(2.0, "base-a-test-cpu")
 
 
 class TestGrammarStats(unittest.TestCase):
diff --git a/test/registered/unit/constrained/test_e2e_constrained_reasoning.py b/test/registered/unit/constrained/test_e2e_constrained_reasoning.py
index 767491c97..6a005202b 100644
--- a/test/registered/unit/constrained/test_e2e_constrained_reasoning.py
+++ b/test/registered/unit/constrained/test_e2e_constrained_reasoning.py
@@ -22,7 +22,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=120, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=120, stage="base-b", runner_config="1-gpu-small")
 
 MODEL = "Qwen/Qwen3-0.6B"
 BASE_URL = "http://127.0.0.1:39877"
diff --git a/test/registered/unit/constrained/test_grammar_manager.py b/test/registered/unit/constrained/test_grammar_manager.py
index 2d25a4bc4..e7605ee44 100644
--- a/test/registered/unit/constrained/test_grammar_manager.py
+++ b/test/registered/unit/constrained/test_grammar_manager.py
@@ -27,7 +27,7 @@ from sglang.srt.constrained.grammar_manager import GrammarManager
 from sglang.srt.constrained.reasoner_grammar_backend import ReasonerGrammarObject
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(2.0, "stage-a-test-cpu")
+register_cpu_ci(2.0, "base-a-test-cpu")
 
 
 def _make_scheduler(grammar_backend_name="none", skip_tokenizer=False):
diff --git a/test/registered/unit/constrained/test_reasoner_grammar_backend.py b/test/registered/unit/constrained/test_reasoner_grammar_backend.py
index 365d42c3c..48b8a79f8 100644
--- a/test/registered/unit/constrained/test_reasoner_grammar_backend.py
+++ b/test/registered/unit/constrained/test_reasoner_grammar_backend.py
@@ -15,7 +15,7 @@ from sglang.srt.constrained.torch_ops.token_filter_torch_ops import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(2.0, "stage-a-test-cpu")
+register_cpu_ci(2.0, "base-a-test-cpu")
 
 
 class _DummyTokenizer:
diff --git a/test/registered/unit/constrained/test_token_filter_ops.py b/test/registered/unit/constrained/test_token_filter_ops.py
index 08dba0b4b..363f6a818 100644
--- a/test/registered/unit/constrained/test_token_filter_ops.py
+++ b/test/registered/unit/constrained/test_token_filter_ops.py
@@ -14,7 +14,7 @@ from sglang.srt.constrained.torch_ops.token_filter_torch_ops import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(2.0, "stage-a-test-cpu")
+register_cpu_ci(2.0, "base-a-test-cpu")
 
 # Conditionally import Triton path
 _has_cuda = torch.cuda.is_available()
diff --git a/test/registered/unit/constrained/test_utils.py b/test/registered/unit/constrained/test_utils.py
index 7384d0971..51ca4470f 100644
--- a/test/registered/unit/constrained/test_utils.py
+++ b/test/registered/unit/constrained/test_utils.py
@@ -14,7 +14,7 @@ import unittest
 from sglang.srt.constrained.utils import is_legacy_structural_tag
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(1.0, "stage-a-test-cpu")
+register_cpu_ci(1.0, "base-a-test-cpu")
 
 
 class TestIsLegacyStructuralTag(unittest.TestCase):
diff --git a/test/registered/unit/disaggregation/test_disaggregation_wire.py b/test/registered/unit/disaggregation/test_disaggregation_wire.py
index 2c635b1d6..334b3121d 100644
--- a/test/registered/unit/disaggregation/test_disaggregation_wire.py
+++ b/test/registered/unit/disaggregation/test_disaggregation_wire.py
@@ -10,7 +10,7 @@ from sglang.srt.disaggregation.common.utils import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=2, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=2, suite="base-a-test-cpu")
 
 
 class TestDisaggregationWire(unittest.TestCase):
diff --git a/test/registered/unit/disaggregation/test_register_to_bootstrap.py b/test/registered/unit/disaggregation/test_register_to_bootstrap.py
index b16574812..2c39681e0 100644
--- a/test/registered/unit/disaggregation/test_register_to_bootstrap.py
+++ b/test/registered/unit/disaggregation/test_register_to_bootstrap.py
@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=5, suite="base-a-test-cpu")
 
 import unittest
 from unittest.mock import MagicMock, call, patch
diff --git a/test/registered/unit/entrypoints/openai/test_matched_stop.py b/test/registered/unit/entrypoints/openai/test_matched_stop.py
index 9c0f5d5ee..9e3e950e0 100644
--- a/test/registered/unit/entrypoints/openai/test_matched_stop.py
+++ b/test/registered/unit/entrypoints/openai/test_matched_stop.py
@@ -3,7 +3,7 @@ import unittest
 from sglang.srt.sampling.sampling_params import MAX_LEN, get_max_seq_length
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 
 class TestRegexPatternMaxLength(unittest.TestCase):
diff --git a/test/registered/unit/entrypoints/openai/test_protocol.py b/test/registered/unit/entrypoints/openai/test_protocol.py
index 776b0aedc..da27ef9ed 100644
--- a/test/registered/unit/entrypoints/openai/test_protocol.py
+++ b/test/registered/unit/entrypoints/openai/test_protocol.py
@@ -32,7 +32,7 @@ from sglang.srt.entrypoints.openai.protocol import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 
 class TestModelCard(unittest.TestCase):
diff --git a/test/registered/unit/entrypoints/openai/test_serving_chat.py b/test/registered/unit/entrypoints/openai/test_serving_chat.py
index 608d2e76b..dbf4ed3f8 100644
--- a/test/registered/unit/entrypoints/openai/test_serving_chat.py
+++ b/test/registered/unit/entrypoints/openai/test_serving_chat.py
@@ -32,7 +32,7 @@ from sglang.srt.managers.template_detection import ReasoningToggleConfig
 from sglang.srt.utils import get_or_create_event_loop
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=11, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=11, suite="base-a-test-cpu")
 
 
 class _MockTokenizerManager:
diff --git a/test/registered/unit/entrypoints/openai/test_serving_completions.py b/test/registered/unit/entrypoints/openai/test_serving_completions.py
index 8f79b14a0..4817f4719 100644
--- a/test/registered/unit/entrypoints/openai/test_serving_completions.py
+++ b/test/registered/unit/entrypoints/openai/test_serving_completions.py
@@ -22,7 +22,7 @@ from sglang.srt.managers.tokenizer_manager import TokenizerManager
 from sglang.srt.utils import get_or_create_event_loop
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=11, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=11, suite="base-a-test-cpu")
 
 
 class _MockTemplateManager:
diff --git a/test/registered/unit/entrypoints/openai/test_serving_embedding.py b/test/registered/unit/entrypoints/openai/test_serving_embedding.py
index 9b8704e32..02c4cbe17 100644
--- a/test/registered/unit/entrypoints/openai/test_serving_embedding.py
+++ b/test/registered/unit/entrypoints/openai/test_serving_embedding.py
@@ -56,7 +56,7 @@ from sglang.srt.entrypoints.openai.serving_embedding import OpenAIServingEmbeddi
 from sglang.srt.managers.io_struct import EmbeddingReqInput
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 
 # Mock TokenizerManager for embedding tests
diff --git a/test/registered/unit/entrypoints/openai/test_serving_transcription.py b/test/registered/unit/entrypoints/openai/test_serving_transcription.py
index b096c6e9c..49dbafc47 100644
--- a/test/registered/unit/entrypoints/openai/test_serving_transcription.py
+++ b/test/registered/unit/entrypoints/openai/test_serving_transcription.py
@@ -28,7 +28,7 @@ from sglang.srt.utils import get_or_create_event_loop
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=4, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=4, suite="base-a-test-cpu")
 
 
 def _chunk(text: str, finish: str = None) -> dict:
diff --git a/test/registered/unit/entrypoints/openai/test_whisper_adapter.py b/test/registered/unit/entrypoints/openai/test_whisper_adapter.py
index 33e353397..1952b290a 100644
--- a/test/registered/unit/entrypoints/openai/test_whisper_adapter.py
+++ b/test/registered/unit/entrypoints/openai/test_whisper_adapter.py
@@ -21,7 +21,7 @@ from sglang.srt.entrypoints.openai.transcription_adapters.whisper import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=2, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=2, suite="base-a-test-cpu")
 
 
 class TestWhisperParseFusedOutput(CustomTestCase):
diff --git a/test/registered/unit/entrypoints/test_ssl_cert_refresher.py b/test/registered/unit/entrypoints/test_ssl_cert_refresher.py
index 119a08348..5b8f88370 100644
--- a/test/registered/unit/entrypoints/test_ssl_cert_refresher.py
+++ b/test/registered/unit/entrypoints/test_ssl_cert_refresher.py
@@ -8,7 +8,7 @@ from sglang.srt.entrypoints.ssl_utils import SSLCertRefresher
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=14, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=14, suite="base-a-test-cpu")
 
 
 def _make_temp_pem(content: bytes) -> str:
diff --git a/test/registered/unit/entrypoints/test_v1_loads_aggregate.py b/test/registered/unit/entrypoints/test_v1_loads_aggregate.py
index 4273f9687..3b4e8b34e 100644
--- a/test/registered/unit/entrypoints/test_v1_loads_aggregate.py
+++ b/test/registered/unit/entrypoints/test_v1_loads_aggregate.py
@@ -11,7 +11,7 @@ from sglang.srt.entrypoints.v1_loads import _compute_aggregate
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 
 def _load(
diff --git a/test/registered/unit/eplb/test_balanced_packing.py b/test/registered/unit/eplb/test_balanced_packing.py
index efa94f94d..a909814e5 100644
--- a/test/registered/unit/eplb/test_balanced_packing.py
+++ b/test/registered/unit/eplb/test_balanced_packing.py
@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 import unittest
 
diff --git a/test/registered/unit/eplb/test_compute_logical_to_rank_dispatch_physical_map.py b/test/registered/unit/eplb/test_compute_logical_to_rank_dispatch_physical_map.py
index f4981060c..5ae2b25da 100644
--- a/test/registered/unit/eplb/test_compute_logical_to_rank_dispatch_physical_map.py
+++ b/test/registered/unit/eplb/test_compute_logical_to_rank_dispatch_physical_map.py
@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 import types
 import unittest
diff --git a/test/registered/unit/function_call/test_function_call_parser.py b/test/registered/unit/function_call/test_function_call_parser.py
index 5e5459ded..9390ca4a8 100644
--- a/test/registered/unit/function_call/test_function_call_parser.py
+++ b/test/registered/unit/function_call/test_function_call_parser.py
@@ -31,7 +31,7 @@ from sglang.srt.function_call.pythonic_detector import PythonicDetector
 from sglang.srt.function_call.qwen3_coder_detector import Qwen3CoderDetector
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=15, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=15, suite="base-a-test-cpu")
 
 
 class TestPythonicDetector(unittest.TestCase):
diff --git a/test/registered/unit/function_call/test_hermes_detector.py b/test/registered/unit/function_call/test_hermes_detector.py
index d1a67c3ae..b6afb16bf 100644
--- a/test/registered/unit/function_call/test_hermes_detector.py
+++ b/test/registered/unit/function_call/test_hermes_detector.py
@@ -7,7 +7,7 @@ from sglang.srt.function_call.hermes_detector import HermesDetector
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(1.0, "stage-a-test-cpu")
+register_cpu_ci(1.0, "base-a-test-cpu")
 
 
 class TestHermesDetector(CustomTestCase):
diff --git a/test/registered/unit/function_call/test_hunyuan_detector.py b/test/registered/unit/function_call/test_hunyuan_detector.py
index 5b64aac8b..e3d9db6d1 100644
--- a/test/registered/unit/function_call/test_hunyuan_detector.py
+++ b/test/registered/unit/function_call/test_hunyuan_detector.py
@@ -8,7 +8,7 @@ from sglang.srt.function_call.hunyuan_detector import HunyuanDetector
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 
 def _make_tools():
diff --git a/test/registered/unit/function_call/test_json_schema_constraint.py b/test/registered/unit/function_call/test_json_schema_constraint.py
index f66943718..91867709c 100644
--- a/test/registered/unit/function_call/test_json_schema_constraint.py
+++ b/test/registered/unit/function_call/test_json_schema_constraint.py
@@ -18,7 +18,7 @@ from sglang.srt.function_call.utils import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(5, "stage-a-test-cpu")
+register_cpu_ci(5, "base-a-test-cpu")
 
 
 class TestJsonSchemaConstraint(unittest.TestCase):
diff --git a/test/registered/unit/function_call/test_llama32_detector.py b/test/registered/unit/function_call/test_llama32_detector.py
index db828ef04..8a5a3f19c 100644
--- a/test/registered/unit/function_call/test_llama32_detector.py
+++ b/test/registered/unit/function_call/test_llama32_detector.py
@@ -7,7 +7,7 @@ from sglang.srt.function_call.llama32_detector import Llama32Detector
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(1.0, "stage-a-test-cpu")
+register_cpu_ci(1.0, "base-a-test-cpu")
 
 
 class TestLlama32Detector(CustomTestCase):
diff --git a/test/registered/unit/function_call/test_mistral_detector.py b/test/registered/unit/function_call/test_mistral_detector.py
index ec49efedb..4dcfe00ef 100644
--- a/test/registered/unit/function_call/test_mistral_detector.py
+++ b/test/registered/unit/function_call/test_mistral_detector.py
@@ -7,7 +7,7 @@ from sglang.srt.function_call.mistral_detector import MistralDetector
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(1.0, "stage-a-test-cpu")
+register_cpu_ci(1.0, "base-a-test-cpu")
 
 
 class TestMistralDetector(CustomTestCase):
diff --git a/test/registered/unit/function_call/test_parallel_tool_calls.py b/test/registered/unit/function_call/test_parallel_tool_calls.py
index 2f5af4d9a..f090d1336 100644
--- a/test/registered/unit/function_call/test_parallel_tool_calls.py
+++ b/test/registered/unit/function_call/test_parallel_tool_calls.py
@@ -23,7 +23,7 @@ from sglang.srt.entrypoints.openai.protocol import Function, Tool
 from sglang.srt.function_call.json_array_parser import JsonArrayParser
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(5, "stage-a-test-cpu")
+register_cpu_ci(5, "base-a-test-cpu")
 
 
 class TestParallelToolCalls(unittest.TestCase):
diff --git a/test/registered/unit/function_call/test_poolside_v1_detector.py b/test/registered/unit/function_call/test_poolside_v1_detector.py
index 2370cf479..23998d298 100644
--- a/test/registered/unit/function_call/test_poolside_v1_detector.py
+++ b/test/registered/unit/function_call/test_poolside_v1_detector.py
@@ -8,7 +8,7 @@ from sglang.srt.function_call.poolside_v1_detector import PoolsideV1Detector
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(1.0, "stage-a-test-cpu")
+register_cpu_ci(1.0, "base-a-test-cpu")
 
 
 class TestPoolsideV1Detector(CustomTestCase):
diff --git a/test/registered/unit/function_call/test_unknown_tool_name.py b/test/registered/unit/function_call/test_unknown_tool_name.py
index ce96c698a..4790c8bca 100644
--- a/test/registered/unit/function_call/test_unknown_tool_name.py
+++ b/test/registered/unit/function_call/test_unknown_tool_name.py
@@ -9,7 +9,7 @@ from sglang.srt.function_call.base_format_detector import BaseFormatDetector
 from sglang.srt.function_call.core_types import StreamingParseResult
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(5, "stage-a-test-cpu")
+register_cpu_ci(5, "base-a-test-cpu")
 
 
 class DummyDetector(BaseFormatDetector):
diff --git a/test/registered/unit/hardware_backend/mlx/test_quantization.py b/test/registered/unit/hardware_backend/mlx/test_quantization.py
index f518d80b3..7065a7eb3 100644
--- a/test/registered/unit/hardware_backend/mlx/test_quantization.py
+++ b/test/registered/unit/hardware_backend/mlx/test_quantization.py
@@ -24,7 +24,7 @@ from sglang.test.ci.ci_register import register_cpu_ci
 # On non-Apple-Silicon CI runners the entire TestCase class skips via the
 # @skipUnless guard below, so this registration is the harmless "yes this
 # test exists" signal the registry requires.
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 _IS_APPLE_SILICON = platform.system() == "Darwin" and platform.machine() == "arm64"
 _HAS_MLX = (
diff --git a/test/registered/unit/layers/quantization/test_fp4_kv_cache_quant_method.py b/test/registered/unit/layers/quantization/test_fp4_kv_cache_quant_method.py
index a6ae13fb2..9726d79f2 100644
--- a/test/registered/unit/layers/quantization/test_fp4_kv_cache_quant_method.py
+++ b/test/registered/unit/layers/quantization/test_fp4_kv_cache_quant_method.py
@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=5, suite="base-a-test-cpu")
 
 import unittest
 
diff --git a/test/registered/unit/layers/quantization/test_mxfp4_sm90_cutlass.py b/test/registered/unit/layers/quantization/test_mxfp4_sm90_cutlass.py
index 85ca0fd27..c48270b0f 100644
--- a/test/registered/unit/layers/quantization/test_mxfp4_sm90_cutlass.py
+++ b/test/registered/unit/layers/quantization/test_mxfp4_sm90_cutlass.py
@@ -20,7 +20,7 @@ import torch
 
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=120, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=120, stage="base-b", runner_config="1-gpu-large")
 
 flashinfer_fused_moe = pytest.importorskip("flashinfer.fused_moe")
 
diff --git a/test/registered/unit/layers/test_conv_layer.py b/test/registered/unit/layers/test_conv_layer.py
index 243a304b5..d4934aa1e 100644
--- a/test/registered/unit/layers/test_conv_layer.py
+++ b/test/registered/unit/layers/test_conv_layer.py
@@ -1,6 +1,6 @@
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=7, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=7, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=7, suite="stage-b-test-1-gpu-small-amd")
 
 import unittest
diff --git a/test/registered/unit/layers/test_mamba_state_scatter_triton.py b/test/registered/unit/layers/test_mamba_state_scatter_triton.py
index 863a2ccc0..b02136462 100644
--- a/test/registered/unit/layers/test_mamba_state_scatter_triton.py
+++ b/test/registered/unit/layers/test_mamba_state_scatter_triton.py
@@ -1,6 +1,6 @@
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=7, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=7, stage="base-b", runner_config="1-gpu-small")
 
 import os
 import unittest
diff --git a/test/registered/unit/layers/test_pooler_score_and_pool.py b/test/registered/unit/layers/test_pooler_score_and_pool.py
index 9ce8d33ff..8eb40d6ba 100644
--- a/test/registered/unit/layers/test_pooler_score_and_pool.py
+++ b/test/registered/unit/layers/test_pooler_score_and_pool.py
@@ -19,7 +19,7 @@ from sglang.srt.layers.pooler import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 
 def _make_forward_batch(
diff --git a/test/registered/unit/lora/test_mem_pool_ep_unit.py b/test/registered/unit/lora/test_mem_pool_ep_unit.py
index 024770dff..7398f6087 100644
--- a/test/registered/unit/lora/test_mem_pool_ep_unit.py
+++ b/test/registered/unit/lora/test_mem_pool_ep_unit.py
@@ -16,7 +16,7 @@ Usage:
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 # CPU-only unit test; no CUDA/distributed dependencies.
-register_cuda_ci(est_time=9, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=9, suite="stage-b-test-1-gpu-small-amd")
 
 import types
diff --git a/test/registered/unit/managers/test_dp_budget.py b/test/registered/unit/managers/test_dp_budget.py
index 5d4e66cea..59d35c294 100644
--- a/test/registered/unit/managers/test_dp_budget.py
+++ b/test/registered/unit/managers/test_dp_budget.py
@@ -17,7 +17,7 @@ maybe_stub_sgl_kernel()
 from sglang.srt.managers.data_parallel_controller import DPBudget
 from sglang.srt.managers.io_struct import GetLoadsReqOutput, WatchLoadUpdateReq
 
-register_cpu_ci(est_time=11, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=11, suite="base-a-test-cpu")
 
 
 _BASE_LOAD = GetLoadsReqOutput(
diff --git a/test/registered/unit/managers/test_hisparse_unit.py b/test/registered/unit/managers/test_hisparse_unit.py
index 14ef091ad..ff2c98b34 100644
--- a/test/registered/unit/managers/test_hisparse_unit.py
+++ b/test/registered/unit/managers/test_hisparse_unit.py
@@ -16,7 +16,7 @@ import torch
 from sglang.srt.utils import is_cuda, is_hip, is_npu, is_xpu
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=10, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small")
 
 # ---------------------------------------------------------------------------
 # Test configuration (small-scale for fast CI runs)
diff --git a/test/registered/unit/managers/test_io_struct.py b/test/registered/unit/managers/test_io_struct.py
index 6e2714570..d9a3b86cd 100644
--- a/test/registered/unit/managers/test_io_struct.py
+++ b/test/registered/unit/managers/test_io_struct.py
@@ -9,7 +9,7 @@ from sglang.test.test_utils import (
     CustomTestCase,
 )
 
-register_cuda_ci(est_time=8, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=8, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=8, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/unit/managers/test_prefill_adder.py b/test/registered/unit/managers/test_prefill_adder.py
index 14d4eab70..c5c51ee06 100644
--- a/test/registered/unit/managers/test_prefill_adder.py
+++ b/test/registered/unit/managers/test_prefill_adder.py
@@ -12,7 +12,7 @@ from sglang.srt.server_args import ServerArgs, set_global_server_args_for_schedu
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=9, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=2, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/unit/managers/test_priority_scheduling_disaggregation.py b/test/registered/unit/managers/test_priority_scheduling_disaggregation.py
index 11c40d6e0..172455598 100644
--- a/test/registered/unit/managers/test_priority_scheduling_disaggregation.py
+++ b/test/registered/unit/managers/test_priority_scheduling_disaggregation.py
@@ -14,7 +14,7 @@ from sglang.srt.managers.schedule_batch import FINISH_ABORT  # noqa: E402
 from sglang.srt.managers.scheduler import Scheduler  # noqa: E402
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=5, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=5, stage="base-b", runner_config="1-gpu-small")
 
 
 class TestDisaggregationPriorityQueueing(unittest.TestCase):
diff --git a/test/registered/unit/managers/test_profile_merger_http_api.py b/test/registered/unit/managers/test_profile_merger_http_api.py
index ccb12dae7..e06fb7a79 100644
--- a/test/registered/unit/managers/test_profile_merger_http_api.py
+++ b/test/registered/unit/managers/test_profile_merger_http_api.py
@@ -4,7 +4,7 @@ import unittest
 from sglang.srt.managers.io_struct import ProfileReqInput
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=8, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=8, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=9, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/unit/managers/test_scheduler_chunked_req_gate.py b/test/registered/unit/managers/test_scheduler_chunked_req_gate.py
index 87a6daf7e..3e6a73c88 100644
--- a/test/registered/unit/managers/test_scheduler_chunked_req_gate.py
+++ b/test/registered/unit/managers/test_scheduler_chunked_req_gate.py
@@ -15,7 +15,7 @@ from sglang.srt.managers.schedule_batch import Req
 from sglang.srt.managers.scheduler import Scheduler
 from sglang.srt.mem_cache.chunk_cache import ChunkCache
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 
 def _make_req(
diff --git a/test/registered/unit/managers/test_scheduler_flush_cache.py b/test/registered/unit/managers/test_scheduler_flush_cache.py
index 0b5d5aadd..73a522d26 100644
--- a/test/registered/unit/managers/test_scheduler_flush_cache.py
+++ b/test/registered/unit/managers/test_scheduler_flush_cache.py
@@ -9,7 +9,7 @@ maybe_stub_sgl_kernel()
 from sglang.srt.managers.io_struct import FlushCacheReqInput
 from sglang.srt.managers.scheduler import Scheduler
 
-register_cpu_ci(est_time=14, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=14, suite="base-a-test-cpu")
 
 
 class TestSchedulerFlushCache(unittest.TestCase):
diff --git a/test/registered/unit/managers/test_scheduler_pause_generation.py b/test/registered/unit/managers/test_scheduler_pause_generation.py
index b1059a33e..c6200584c 100644
--- a/test/registered/unit/managers/test_scheduler_pause_generation.py
+++ b/test/registered/unit/managers/test_scheduler_pause_generation.py
@@ -11,7 +11,7 @@ from sglang.srt.managers.io_struct import PauseGenerationReqInput
 from sglang.srt.managers.scheduler import Scheduler
 from sglang.srt.managers.scheduler_runtime_checker_mixin import PoolStats
 
-register_cpu_ci(est_time=15, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=15, suite="base-a-test-cpu")
 
 
 class TestSchedulerPauseGeneration(unittest.TestCase):
diff --git a/test/registered/unit/managers/test_template_manager.py b/test/registered/unit/managers/test_template_manager.py
index 18d01e11b..dd4db0c41 100644
--- a/test/registered/unit/managers/test_template_manager.py
+++ b/test/registered/unit/managers/test_template_manager.py
@@ -10,7 +10,7 @@ from sglang.srt.managers.template_detection import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(2.0, "stage-a-test-cpu")
+register_cpu_ci(2.0, "base-a-test-cpu")
 
 
 class _DummyTokenizer:
diff --git a/test/registered/unit/mem_cache/test_decode_radix_lock_ref.py b/test/registered/unit/mem_cache/test_decode_radix_lock_ref.py
index f392e9fcd..863942000 100644
--- a/test/registered/unit/mem_cache/test_decode_radix_lock_ref.py
+++ b/test/registered/unit/mem_cache/test_decode_radix_lock_ref.py
@@ -22,7 +22,7 @@ Usage:
 
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=10, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=10, suite="stage-b-test-1-gpu-small-amd")
 
 import unittest
diff --git a/test/registered/unit/mem_cache/test_evict_policy.py b/test/registered/unit/mem_cache/test_evict_policy.py
index 8365c0580..643dc0bd7 100644
--- a/test/registered/unit/mem_cache/test_evict_policy.py
+++ b/test/registered/unit/mem_cache/test_evict_policy.py
@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 import unittest
 from unittest.mock import MagicMock
diff --git a/test/registered/unit/mem_cache/test_mamba_unittest.py b/test/registered/unit/mem_cache/test_mamba_unittest.py
index 4cc253e56..f5ff1d2f1 100755
--- a/test/registered/unit/mem_cache/test_mamba_unittest.py
+++ b/test/registered/unit/mem_cache/test_mamba_unittest.py
@@ -23,7 +23,7 @@ from sglang.srt.server_args import ServerArgs, set_global_server_args_for_schedu
 from sglang.srt.utils import get_device
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=10, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=9, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/unit/mem_cache/test_nsa_pool_host_unit.py b/test/registered/unit/mem_cache/test_nsa_pool_host_unit.py
index c4fafdb56..17fe9d691 100644
--- a/test/registered/unit/mem_cache/test_nsa_pool_host_unit.py
+++ b/test/registered/unit/mem_cache/test_nsa_pool_host_unit.py
@@ -12,7 +12,7 @@ from sglang.srt.mem_cache.memory_pool_host import (
 from sglang.srt.utils import is_cuda, is_hip, is_npu, is_xpu
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=9, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-small")
 
 
 class TestNSAHiCacheTransfer(unittest.TestCase):
diff --git a/test/registered/unit/mem_cache/test_radix_cache_slru_accuracy.py b/test/registered/unit/mem_cache/test_radix_cache_slru_accuracy.py
index a167b4f02..44d4e0419 100644
--- a/test/registered/unit/mem_cache/test_radix_cache_slru_accuracy.py
+++ b/test/registered/unit/mem_cache/test_radix_cache_slru_accuracy.py
@@ -13,7 +13,7 @@ from sglang.srt.mem_cache.memory_pool import MHATokenToKVPool, ReqToTokenPool
 from sglang.srt.mem_cache.radix_cache import RadixCache, RadixKey
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=8, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=8, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=8, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/unit/mem_cache/test_radix_cache_unit.py b/test/registered/unit/mem_cache/test_radix_cache_unit.py
index 9cbb7bc77..a445b9b3d 100644
--- a/test/registered/unit/mem_cache/test_radix_cache_unit.py
+++ b/test/registered/unit/mem_cache/test_radix_cache_unit.py
@@ -21,7 +21,7 @@ from sglang.srt.mem_cache.common import available_and_evictable_str
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 # CPU-based unit test, runs quickly on any GPU runner
-register_cuda_ci(est_time=15, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=15, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=5, suite="stage-b-test-1-gpu-small-amd")
 
 import random
diff --git a/test/registered/unit/mem_cache/test_radix_force_miss.py b/test/registered/unit/mem_cache/test_radix_force_miss.py
index 24d5f345b..c91481bba 100644
--- a/test/registered/unit/mem_cache/test_radix_force_miss.py
+++ b/test/registered/unit/mem_cache/test_radix_force_miss.py
@@ -7,7 +7,7 @@ RadixCache.
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=5, suite="base-a-test-cpu")
 
 import unittest
 import unittest.mock
diff --git a/test/registered/unit/mem_cache/test_streaming_session_unit.py b/test/registered/unit/mem_cache/test_streaming_session_unit.py
index f833872fb..b0fe8cd9e 100644
--- a/test/registered/unit/mem_cache/test_streaming_session_unit.py
+++ b/test/registered/unit/mem_cache/test_streaming_session_unit.py
@@ -7,7 +7,7 @@ from sglang.srt.mem_cache.base_prefix_cache import MatchResult
 from sglang.srt.session.streaming_session import SessionSlot, StreamingSession
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=12, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=12, suite="base-a-test-cpu")
 
 
 class _FakeAllocator:
diff --git a/test/registered/unit/mem_cache/test_swa_alloc_extend_page_estimation.py b/test/registered/unit/mem_cache/test_swa_alloc_extend_page_estimation.py
index c1af18e4f..15ff77ea3 100644
--- a/test/registered/unit/mem_cache/test_swa_alloc_extend_page_estimation.py
+++ b/test/registered/unit/mem_cache/test_swa_alloc_extend_page_estimation.py
@@ -15,7 +15,7 @@ from sglang.srt.mem_cache.swa_memory_pool import SWATokenToKVPoolAllocator
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=2, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=2, suite="base-a-test-cpu")
 
 
 def _make_self(*, page_size: int, full_available: int, swa_available: int):
diff --git a/test/registered/unit/mem_cache/test_swa_eviction_boundary.py b/test/registered/unit/mem_cache/test_swa_eviction_boundary.py
index 3dd0bf2f2..ef643c8bb 100644
--- a/test/registered/unit/mem_cache/test_swa_eviction_boundary.py
+++ b/test/registered/unit/mem_cache/test_swa_eviction_boundary.py
@@ -26,7 +26,7 @@ from sglang.srt.mem_cache.swa_radix_cache import SWARadixCache
 from sglang.srt.utils import get_device
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=12, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=12, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=10, suite="stage-b-test-1-gpu-small-amd")
 
 # ---------------------------------------------------------------------------
diff --git a/test/registered/unit/mem_cache/test_swa_lock_release_lifecycle.py b/test/registered/unit/mem_cache/test_swa_lock_release_lifecycle.py
index 7409fab82..173f338dc 100644
--- a/test/registered/unit/mem_cache/test_swa_lock_release_lifecycle.py
+++ b/test/registered/unit/mem_cache/test_swa_lock_release_lifecycle.py
@@ -30,7 +30,7 @@ from sglang.srt.utils import get_device
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=12, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=12, stage="base-b", runner_config="1-gpu-small")
 
 
 def _build_tree(
diff --git a/test/registered/unit/mem_cache/test_swa_unittest.py b/test/registered/unit/mem_cache/test_swa_unittest.py
index 08f9a0c1c..96e0bd1db 100644
--- a/test/registered/unit/mem_cache/test_swa_unittest.py
+++ b/test/registered/unit/mem_cache/test_swa_unittest.py
@@ -21,7 +21,7 @@ from sglang.srt.utils import get_device
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=9, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=10, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/unit/mem_cache/test_unified_radix_cache_bench.py b/test/registered/unit/mem_cache/test_unified_radix_cache_bench.py
index b76ecbba4..bef9eec69 100644
--- a/test/registered/unit/mem_cache/test_unified_radix_cache_bench.py
+++ b/test/registered/unit/mem_cache/test_unified_radix_cache_bench.py
@@ -38,7 +38,7 @@ from sglang.srt.server_args import ServerArgs, set_global_server_args_for_schedu
 from sglang.srt.utils import get_device
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=25, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=25, stage="base-b", runner_config="1-gpu-small")
 
 # ---------------------------------------------------------------------------
 # Constants
diff --git a/test/registered/unit/mem_cache/test_unified_radix_cache_unittest.py b/test/registered/unit/mem_cache/test_unified_radix_cache_unittest.py
index 7074084d9..43202bc27 100644
--- a/test/registered/unit/mem_cache/test_unified_radix_cache_unittest.py
+++ b/test/registered/unit/mem_cache/test_unified_radix_cache_unittest.py
@@ -49,7 +49,7 @@ from sglang.srt.utils import get_device
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=10, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=10, stage="base-b", runner_config="1-gpu-small")
 
 
 @dataclass(frozen=True)
diff --git a/test/registered/unit/model_executor/test_pool_configurator.py b/test/registered/unit/model_executor/test_pool_configurator.py
index ea98b2c4d..761efeb88 100644
--- a/test/registered/unit/model_executor/test_pool_configurator.py
+++ b/test/registered/unit/model_executor/test_pool_configurator.py
@@ -12,7 +12,7 @@ from unittest.mock import MagicMock, patch
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 
 @contextlib.contextmanager
diff --git a/test/registered/unit/model_loader/test_modelopt_export.py b/test/registered/unit/model_loader/test_modelopt_export.py
index 22421c24b..a530f1745 100644
--- a/test/registered/unit/model_loader/test_modelopt_export.py
+++ b/test/registered/unit/model_loader/test_modelopt_export.py
@@ -19,7 +19,7 @@ from sglang.srt.configs.model_config import ModelConfig
 from sglang.srt.model_loader.loader import ModelOptModelLoader
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=11, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=9, suite="stage-b-test-1-gpu-small-amd")
 
 # Note: PYTHONPATH=python should be set when running tests
diff --git a/test/registered/unit/model_loader/test_modelopt_loader.py b/test/registered/unit/model_loader/test_modelopt_loader.py
index 9acdfaabb..fe3724b67 100644
--- a/test/registered/unit/model_loader/test_modelopt_loader.py
+++ b/test/registered/unit/model_loader/test_modelopt_loader.py
@@ -30,7 +30,7 @@ CALIBRATION_BATCH_SIZE = 36
 CALIBRATION_NUM_SAMPLES = 512
 DEFAULT_DEVICE = "cuda:0"
 
-register_cuda_ci(est_time=11, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-small")
 
 
 class TestModelOptModelLoader(CustomTestCase):
diff --git a/test/registered/unit/model_loader/test_prefetch_checkpoints.py b/test/registered/unit/model_loader/test_prefetch_checkpoints.py
index 0a3e91faf..e417e60b3 100644
--- a/test/registered/unit/model_loader/test_prefetch_checkpoints.py
+++ b/test/registered/unit/model_loader/test_prefetch_checkpoints.py
@@ -18,7 +18,7 @@ from sglang.srt.model_loader.weight_utils import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 
 class TestPrefetchWeightsIdentical(unittest.TestCase):
diff --git a/test/registered/unit/model_loader/test_runai_model_streamer_loader.py b/test/registered/unit/model_loader/test_runai_model_streamer_loader.py
index 8b409c560..e99665c98 100644
--- a/test/registered/unit/model_loader/test_runai_model_streamer_loader.py
+++ b/test/registered/unit/model_loader/test_runai_model_streamer_loader.py
@@ -15,7 +15,7 @@ from sglang.srt.models.deepseek_common import deepseek_weight_loader
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 
 class _FakeModel:
diff --git a/test/registered/unit/models/test_llava.py b/test/registered/unit/models/test_llava.py
index 7a211193f..58f4a5c59 100644
--- a/test/registered/unit/models/test_llava.py
+++ b/test/registered/unit/models/test_llava.py
@@ -5,7 +5,7 @@ from sglang.srt.models.llava import AutoModel, LlavaForConditionalGeneration
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=9, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=9, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/unit/models/test_nemotron_h_weight_loading.py b/test/registered/unit/models/test_nemotron_h_weight_loading.py
index 8a5e30a4e..0f886cdda 100644
--- a/test/registered/unit/models/test_nemotron_h_weight_loading.py
+++ b/test/registered/unit/models/test_nemotron_h_weight_loading.py
@@ -7,7 +7,7 @@ parameters absent from the current runtime model.
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=4, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=4, suite="base-a-test-cpu")
 
 import unittest
 from types import SimpleNamespace
diff --git a/test/registered/unit/models/test_qwen3_5_packed_weight_loader.py b/test/registered/unit/models/test_qwen3_5_packed_weight_loader.py
index f2d60debb..e384e1e95 100644
--- a/test/registered/unit/models/test_qwen3_5_packed_weight_loader.py
+++ b/test/registered/unit/models/test_qwen3_5_packed_weight_loader.py
@@ -10,7 +10,7 @@ Regression test for https://github.com/sgl-project/sglang/issues/23051
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=4, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=4, suite="base-a-test-cpu")
 
 import unittest
 from types import SimpleNamespace
diff --git a/test/registered/unit/observability/test_cpu_monitor.py b/test/registered/unit/observability/test_cpu_monitor.py
index 3cb4e3bd5..35eb29771 100644
--- a/test/registered/unit/observability/test_cpu_monitor.py
+++ b/test/registered/unit/observability/test_cpu_monitor.py
@@ -6,7 +6,7 @@ from unittest.mock import MagicMock, patch
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=60, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=60, suite="base-a-test-cpu", nightly=True)
 
 
 class TestCpuMonitor(unittest.TestCase):
diff --git a/test/registered/unit/observability/test_forward_pass_metrics.py b/test/registered/unit/observability/test_forward_pass_metrics.py
index d63c89993..a34470951 100644
--- a/test/registered/unit/observability/test_forward_pass_metrics.py
+++ b/test/registered/unit/observability/test_forward_pass_metrics.py
@@ -1,6 +1,6 @@
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=5, suite="base-a-test-cpu")
 
 import types
 import unittest
diff --git a/test/registered/unit/observability/test_func_timer.py b/test/registered/unit/observability/test_func_timer.py
index 30608ca8e..8efd5a773 100644
--- a/test/registered/unit/observability/test_func_timer.py
+++ b/test/registered/unit/observability/test_func_timer.py
@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 import asyncio
 import unittest
diff --git a/test/registered/unit/observability/test_label_transform.py b/test/registered/unit/observability/test_label_transform.py
index f71c66bd1..321ed6455 100644
--- a/test/registered/unit/observability/test_label_transform.py
+++ b/test/registered/unit/observability/test_label_transform.py
@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 import unittest
 
diff --git a/test/registered/unit/observability/test_metrics_utils.py b/test/registered/unit/observability/test_metrics_utils.py
index 6eb26d8ff..0789b120e 100644
--- a/test/registered/unit/observability/test_metrics_utils.py
+++ b/test/registered/unit/observability/test_metrics_utils.py
@@ -6,7 +6,7 @@ from sglang.srt.observability.utils import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 
 class TestMetricsUtils(unittest.TestCase):
diff --git a/test/registered/unit/observability/test_request_metrics_exporter.py b/test/registered/unit/observability/test_request_metrics_exporter.py
index 683bb10b6..d481032ae 100644
--- a/test/registered/unit/observability/test_request_metrics_exporter.py
+++ b/test/registered/unit/observability/test_request_metrics_exporter.py
@@ -5,7 +5,7 @@ from typing import Any, Dict, List, Optional
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 import asyncio
 import json
diff --git a/test/registered/unit/observability/test_startup_func_log_and_timer.py b/test/registered/unit/observability/test_startup_func_log_and_timer.py
index 90fcf5074..2eae0172b 100644
--- a/test/registered/unit/observability/test_startup_func_log_and_timer.py
+++ b/test/registered/unit/observability/test_startup_func_log_and_timer.py
@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 import unittest
 from unittest.mock import MagicMock, patch
diff --git a/test/registered/unit/observability/test_trace.py b/test/registered/unit/observability/test_trace.py
index 6a03324fc..18620e4bf 100644
--- a/test/registered/unit/observability/test_trace.py
+++ b/test/registered/unit/observability/test_trace.py
@@ -4,7 +4,7 @@ import os
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 import threading
 import unittest
diff --git a/test/registered/unit/parser/test_code_completion_parser.py b/test/registered/unit/parser/test_code_completion_parser.py
index b8a411585..334e89476 100644
--- a/test/registered/unit/parser/test_code_completion_parser.py
+++ b/test/registered/unit/parser/test_code_completion_parser.py
@@ -18,7 +18,7 @@ from sglang.srt.parser.code_completion_parser import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 
 class TestFimPosition(CustomTestCase):
diff --git a/test/registered/unit/parser/test_conversation.py b/test/registered/unit/parser/test_conversation.py
index 72bfb8cca..3cb8a65d0 100644
--- a/test/registered/unit/parser/test_conversation.py
+++ b/test/registered/unit/parser/test_conversation.py
@@ -32,7 +32,7 @@ from sglang.srt.parser.conversation import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 
 class TestConversationGetPrompt(CustomTestCase):
diff --git a/test/registered/unit/parser/test_harmony_parser.py b/test/registered/unit/parser/test_harmony_parser.py
index 26242a9db..da22fa4bb 100644
--- a/test/registered/unit/parser/test_harmony_parser.py
+++ b/test/registered/unit/parser/test_harmony_parser.py
@@ -14,7 +14,7 @@ from sglang.srt.parser.harmony_parser import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 
 class TestEvent(CustomTestCase):
diff --git a/test/registered/unit/parser/test_jinja_template_utils.py b/test/registered/unit/parser/test_jinja_template_utils.py
index 3396de98a..791c53446 100644
--- a/test/registered/unit/parser/test_jinja_template_utils.py
+++ b/test/registered/unit/parser/test_jinja_template_utils.py
@@ -9,7 +9,7 @@ from sglang.srt.parser.jinja_template_utils import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 
 class TestTemplateContentFormatDetection(CustomTestCase):
diff --git a/test/registered/unit/parser/test_reasoning_content_without_parser.py b/test/registered/unit/parser/test_reasoning_content_without_parser.py
index f8eada591..20bb9c221 100644
--- a/test/registered/unit/parser/test_reasoning_content_without_parser.py
+++ b/test/registered/unit/parser/test_reasoning_content_without_parser.py
@@ -4,7 +4,7 @@ from sglang.srt.parser.reasoning_parser import ReasoningParser
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 # Simulated model output that contains think tags (e.g. from DeepSeek-R1)
 THINK_OUTPUT = (
diff --git a/test/registered/unit/parser/test_reasoning_parser.py b/test/registered/unit/parser/test_reasoning_parser.py
index 85fff23aa..3d9ea07f7 100644
--- a/test/registered/unit/parser/test_reasoning_parser.py
+++ b/test/registered/unit/parser/test_reasoning_parser.py
@@ -18,7 +18,7 @@ from sglang.srt.parser.reasoning_parser import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 
 class TestStreamingParseResult(CustomTestCase):
diff --git a/test/registered/unit/platforms/test_platform_interface.py b/test/registered/unit/platforms/test_platform_interface.py
index 8c8cec606..a4b4cd9f7 100644
--- a/test/registered/unit/platforms/test_platform_interface.py
+++ b/test/registered/unit/platforms/test_platform_interface.py
@@ -21,7 +21,7 @@ from sglang.srt.platforms.interface import SRTPlatform
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 
 # ---------------------------------------------------------------------------
diff --git a/test/registered/unit/plugins/test_hook_registry.py b/test/registered/unit/plugins/test_hook_registry.py
index b01ed3b37..e898735e1 100644
--- a/test/registered/unit/plugins/test_hook_registry.py
+++ b/test/registered/unit/plugins/test_hook_registry.py
@@ -16,7 +16,7 @@ from sglang.srt.plugins.hook_registry import HookRegistry, HookType, plugin_hook
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 # ---------------------------------------------------------------------------
 # Helpers: synthetic module creation
diff --git a/test/registered/unit/plugins/test_load_plugins.py b/test/registered/unit/plugins/test_load_plugins.py
index be2e13570..af4774b95 100644
--- a/test/registered/unit/plugins/test_load_plugins.py
+++ b/test/registered/unit/plugins/test_load_plugins.py
@@ -19,7 +19,7 @@ from sglang.srt.plugins import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 
 def _make_ep(name, dist_name=None, load_fn=None):
diff --git a/test/registered/unit/sampling/test_custom_logit_processor.py b/test/registered/unit/sampling/test_custom_logit_processor.py
index eb9a76d0c..d02e50f40 100644
--- a/test/registered/unit/sampling/test_custom_logit_processor.py
+++ b/test/registered/unit/sampling/test_custom_logit_processor.py
@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 import json
 import unittest
diff --git a/test/registered/unit/sampling/test_penaltylib.py b/test/registered/unit/sampling/test_penaltylib.py
index 5e39ed32c..77130044e 100644
--- a/test/registered/unit/sampling/test_penaltylib.py
+++ b/test/registered/unit/sampling/test_penaltylib.py
@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=9, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=9, suite="base-a-test-cpu")
 
 import unittest
 from unittest.mock import MagicMock
diff --git a/test/registered/unit/sampling/test_sampling_batch_info.py b/test/registered/unit/sampling/test_sampling_batch_info.py
index c3df29892..7bbdfa8a9 100644
--- a/test/registered/unit/sampling/test_sampling_batch_info.py
+++ b/test/registered/unit/sampling/test_sampling_batch_info.py
@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=9, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=9, suite="base-a-test-cpu")
 
 import unittest
 from unittest.mock import MagicMock, patch
diff --git a/test/registered/unit/sampling/test_sampling_params.py b/test/registered/unit/sampling/test_sampling_params.py
index 67c51a1a2..c245c8c72 100644
--- a/test/registered/unit/sampling/test_sampling_params.py
+++ b/test/registered/unit/sampling/test_sampling_params.py
@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 import unittest
 from unittest.mock import MagicMock
diff --git a/test/registered/unit/server_args/test_server_args.py b/test/registered/unit/server_args/test_server_args.py
index 939a91b49..69fc94dd8 100644
--- a/test/registered/unit/server_args/test_server_args.py
+++ b/test/registered/unit/server_args/test_server_args.py
@@ -10,7 +10,7 @@ from sglang.test.test_utils import (
     CustomTestCase,
 )
 
-register_cpu_ci(est_time=10, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
 
 # Mock get_device() so all tests run on CPU-only CI runners
 _mock_device = patch("sglang.srt.server_args.get_device", return_value="cuda")
diff --git a/test/registered/unit/spec/test_adaptive_spec_params.py b/test/registered/unit/spec/test_adaptive_spec_params.py
index 41fe14954..b1f7937d3 100644
--- a/test/registered/unit/spec/test_adaptive_spec_params.py
+++ b/test/registered/unit/spec/test_adaptive_spec_params.py
@@ -3,7 +3,7 @@ import unittest
 from sglang.srt.speculative.adaptive_spec_params import AdaptiveSpeculativeParams
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 
 class TestAdaptiveSpeculativeParams(unittest.TestCase):
diff --git a/test/registered/unit/spec/test_ngram_corpus.py b/test/registered/unit/spec/test_ngram_corpus.py
index 8d491b5b7..49d44aa3b 100644
--- a/test/registered/unit/spec/test_ngram_corpus.py
+++ b/test/registered/unit/spec/test_ngram_corpus.py
@@ -13,7 +13,7 @@ from sglang.srt.speculative.cpp_ngram.ngram_corpus import NgramCorpus
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=26, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=26, suite="base-a-test-cpu")
 
 
 def _make_corpus(match_type="BFS", **kwargs):
diff --git a/test/registered/unit/spec/test_spec_registry.py b/test/registered/unit/spec/test_spec_registry.py
index ee41e16e0..baeb9d632 100644
--- a/test/registered/unit/spec/test_spec_registry.py
+++ b/test/registered/unit/spec/test_spec_registry.py
@@ -12,7 +12,7 @@ from sglang.srt.speculative.spec_registry import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=5, suite="base-a-test-cpu")
 
 
 class _RegistryIsolated(CustomTestCase):
diff --git a/test/registered/unit/spec/test_spec_utils_traverse_tree.py b/test/registered/unit/spec/test_spec_utils_traverse_tree.py
index d580e024c..69c5c66c2 100644
--- a/test/registered/unit/spec/test_spec_utils_traverse_tree.py
+++ b/test/registered/unit/spec/test_spec_utils_traverse_tree.py
@@ -14,7 +14,7 @@ import torch
 from sglang.srt.speculative.spec_utils import traverse_tree
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=4, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=4, suite="base-a-test-cpu")
 
 
 class TestTraverseTreePassesIntsToGrammar(unittest.TestCase):
diff --git a/test/registered/unit/test_bench_long_context.py b/test/registered/unit/test_bench_long_context.py
index 304cea3d4..0c8b81bde 100644
--- a/test/registered/unit/test_bench_long_context.py
+++ b/test/registered/unit/test_bench_long_context.py
@@ -16,7 +16,7 @@ from unittest.mock import MagicMock, patch
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=5, suite="base-a-test-cpu")
 
 REPO_ROOT = Path(__file__).resolve().parents[3]
 HICACHE_DIR = REPO_ROOT / "benchmark" / "hicache"
diff --git a/test/registered/unit/test_no_bare_pytest_main.py b/test/registered/unit/test_no_bare_pytest_main.py
index 5dda5c34f..8d47f012a 100644
--- a/test/registered/unit/test_no_bare_pytest_main.py
+++ b/test/registered/unit/test_no_bare_pytest_main.py
@@ -5,7 +5,7 @@ import unittest
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=5, suite="base-a-test-cpu")
 
 
 _REPO_ROOT = pathlib.Path(__file__).resolve().parents[3]
diff --git a/test/registered/unit/test_runai_utils.py b/test/registered/unit/test_runai_utils.py
index 83a0adf1a..910559f5e 100644
--- a/test/registered/unit/test_runai_utils.py
+++ b/test/registered/unit/test_runai_utils.py
@@ -6,7 +6,7 @@ from sglang.srt.utils.runai_utils import ObjectStorageModel, is_runai_obj_uri
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 
 class TestRunaiUtils(CustomTestCase):
diff --git a/test/registered/unit/tokenizer/test_tiktoken_tokenizer.py b/test/registered/unit/tokenizer/test_tiktoken_tokenizer.py
index 8b3023317..7fe997bc6 100644
--- a/test/registered/unit/tokenizer/test_tiktoken_tokenizer.py
+++ b/test/registered/unit/tokenizer/test_tiktoken_tokenizer.py
@@ -6,7 +6,7 @@ from unittest.mock import MagicMock, patch
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 from sglang.srt.tokenizer.tiktoken_tokenizer import (
     CONTROL_TOKEN_TEXTS,
diff --git a/test/registered/unit/tools/test_docker_build_metadata_args.py b/test/registered/unit/tools/test_docker_build_metadata_args.py
index c4ee4c042..7aca2444b 100644
--- a/test/registered/unit/tools/test_docker_build_metadata_args.py
+++ b/test/registered/unit/tools/test_docker_build_metadata_args.py
@@ -19,7 +19,7 @@ def _load_module(name, path):
 
 
 register_cpu_ci = _load_module("ci_register", CI_REGISTER_PATH).register_cpu_ci
-register_cpu_ci(est_time=0, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=0, suite="base-a-test-cpu")
 
 
 class TestDockerBuildMetadataArgs(unittest.TestCase):
diff --git a/test/registered/unit/tools/test_get_version_tag.py b/test/registered/unit/tools/test_get_version_tag.py
index 71370ad1e..ed068d18b 100644
--- a/test/registered/unit/tools/test_get_version_tag.py
+++ b/test/registered/unit/tools/test_get_version_tag.py
@@ -33,7 +33,7 @@ def _load_module(name, path):
 
 
 register_cpu_ci = _load_module("ci_register", CI_REGISTER_PATH).register_cpu_ci
-register_cpu_ci(est_time=0, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=0, suite="base-a-test-cpu")
 
 
 class TestGetVersionTag(unittest.TestCase):
diff --git a/test/registered/unit/utils/test_auth.py b/test/registered/unit/utils/test_auth.py
index ac4b9b8e8..e690d8576 100644
--- a/test/registered/unit/utils/test_auth.py
+++ b/test/registered/unit/utils/test_auth.py
@@ -11,7 +11,7 @@ from sglang.srt.utils.auth import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(1.0, "stage-a-test-cpu")
+register_cpu_ci(1.0, "base-a-test-cpu")
 
 
 class TestAuthDecision(CustomTestCase):
diff --git a/test/registered/unit/utils/test_gauge_histogram.py b/test/registered/unit/utils/test_gauge_histogram.py
index b2dac54f9..d4a355716 100644
--- a/test/registered/unit/utils/test_gauge_histogram.py
+++ b/test/registered/unit/utils/test_gauge_histogram.py
@@ -1,6 +1,6 @@
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 import unittest
 
diff --git a/test/registered/unit/utils/test_hf_transformers.py b/test/registered/unit/utils/test_hf_transformers.py
index df7fddb29..b29839d0e 100644
--- a/test/registered/unit/utils/test_hf_transformers.py
+++ b/test/registered/unit/utils/test_hf_transformers.py
@@ -24,7 +24,7 @@ from sglang.srt.utils.hf_transformers.tokenizer import _fix_special_tokens_patte
 from sglang.srt.utils.hf_transformers_patches import normalize_rope_scaling_compat
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 
 # ---------------------------------------------------------------------------
diff --git a/test/registered/unit/utils/test_hf_transformers_fastokens.py b/test/registered/unit/utils/test_hf_transformers_fastokens.py
index 9d63e91d5..0b8017833 100644
--- a/test/registered/unit/utils/test_hf_transformers_fastokens.py
+++ b/test/registered/unit/utils/test_hf_transformers_fastokens.py
@@ -12,7 +12,7 @@ from sglang.test.test_utils import (
 
 TOKENIZER_MODEL = DEFAULT_SMALL_MODEL_NAME_FOR_TEST_QWEN
 
-register_cpu_ci(est_time=30, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=30, suite="base-a-test-cpu")
 
 
 try:
diff --git a/test/registered/unit/utils/test_http_server_auth.py b/test/registered/unit/utils/test_http_server_auth.py
index 69f50154e..f2a42101e 100644
--- a/test/registered/unit/utils/test_http_server_auth.py
+++ b/test/registered/unit/utils/test_http_server_auth.py
@@ -10,7 +10,7 @@ import unittest
 from sglang.srt.utils.auth import AuthLevel, decide_request_auth
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 
 class TestHttpServerAdminAuth(unittest.TestCase):
diff --git a/test/registered/unit/utils/test_json_response.py b/test/registered/unit/utils/test_json_response.py
index 764de5f53..6f83290d0 100644
--- a/test/registered/unit/utils/test_json_response.py
+++ b/test/registered/unit/utils/test_json_response.py
@@ -10,7 +10,7 @@ from sglang.srt.utils.json_response import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 
 class TestJSONResponseUtils(unittest.TestCase):
diff --git a/test/registered/unit/utils/test_patch_tokenizer.py b/test/registered/unit/utils/test_patch_tokenizer.py
index ea0ba7533..52e6be73b 100644
--- a/test/registered/unit/utils/test_patch_tokenizer.py
+++ b/test/registered/unit/utils/test_patch_tokenizer.py
@@ -10,7 +10,7 @@ from sglang.srt.utils.patch_tokenizer import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=30, suite="stage-a-test-cpu", nightly=True)
+register_cpu_ci(est_time=30, suite="base-a-test-cpu", nightly=True)
 
 
 class TestPatchTokenizerEndToEndTest(unittest.TestCase):
diff --git a/test/registered/unit/utils/test_profile_merger.py b/test/registered/unit/utils/test_profile_merger.py
index a7f5d2cfb..879afb514 100644
--- a/test/registered/unit/utils/test_profile_merger.py
+++ b/test/registered/unit/utils/test_profile_merger.py
@@ -17,7 +17,7 @@ from sglang.srt.managers.io_struct import ProfileReq, ProfileReqInput, ProfileRe
 from sglang.srt.utils.profile_merger import ProfileMerger
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=9, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=9, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=8, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/unit/utils/test_subprocess_watchdog.py b/test/registered/unit/utils/test_subprocess_watchdog.py
index 48873ecbe..eb3df663f 100644
--- a/test/registered/unit/utils/test_subprocess_watchdog.py
+++ b/test/registered/unit/utils/test_subprocess_watchdog.py
@@ -24,7 +24,7 @@ from sglang.srt.utils.watchdog import SubprocessWatchdog
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=9, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=9, suite="base-a-test-cpu")
 
 
 def healthy_worker():
diff --git a/test/registered/unit/utils/test_weight_checker.py b/test/registered/unit/utils/test_weight_checker.py
index ab84cfb46..a253e6155 100644
--- a/test/registered/unit/utils/test_weight_checker.py
+++ b/test/registered/unit/utils/test_weight_checker.py
@@ -38,7 +38,7 @@ from sglang.srt.utils.weight_checker import (
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=30, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=30, stage="base-b", runner_config="1-gpu-small")
 
 
 # ---------------------------------------------------------------------------
diff --git a/test/registered/utils/test_log_utils.py b/test/registered/utils/test_log_utils.py
index b65d89017..9ea137e3e 100644
--- a/test/registered/utils/test_log_utils.py
+++ b/test/registered/utils/test_log_utils.py
@@ -10,7 +10,7 @@ from pathlib import Path
 from sglang.srt.utils.log_utils import create_log_targets, log_json
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=6, suite="base-a-test-cpu")
 
 _LOG_PREFIX_RE = re.compile(r"^\[\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2}\] ")
 
diff --git a/test/registered/utils/test_network_address.py b/test/registered/utils/test_network_address.py
index ff05206c2..2ef628433 100644
--- a/test/registered/utils/test_network_address.py
+++ b/test/registered/utils/test_network_address.py
@@ -6,7 +6,7 @@ from sglang.srt.server_args import PortArgs, ServerArgs
 from sglang.srt.utils.network import NetworkAddress
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 # Mock get_device() so ServerArgs tests run on CPU-only CI runners
 _mock_device = patch("sglang.srt.server_args.get_device", return_value="cuda")
diff --git a/test/registered/utils/test_numa_utils.py b/test/registered/utils/test_numa_utils.py
index a66093334..a0863a62d 100644
--- a/test/registered/utils/test_numa_utils.py
+++ b/test/registered/utils/test_numa_utils.py
@@ -9,9 +9,9 @@ from sglang.srt.utils.numa_utils import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci, register_cuda_ci
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
-register_cuda_ci(est_time=10, stage="stage-c", runner_config="4-gpu-gb200")
-register_cuda_ci(est_time=10, stage="stage-c", runner_config="8-gpu-b200")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
+register_cuda_ci(est_time=10, stage="base-c", runner_config="4-gpu-gb200")
+register_cuda_ci(est_time=10, stage="base-c", runner_config="8-gpu-b200")
 
 
 class TestIsNumaAvailable(unittest.TestCase):
diff --git a/test/registered/utils/test_socket_utils.py b/test/registered/utils/test_socket_utils.py
index 63a42ac39..6e1d4b95c 100644
--- a/test/registered/utils/test_socket_utils.py
+++ b/test/registered/utils/test_socket_utils.py
@@ -15,7 +15,7 @@ from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 from sglang.utils import normalize_base_url, release_port, reserve_port
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 
 class TestTryBindSocket(CustomTestCase):
diff --git a/test/registered/vlm/test_evs.py b/test/registered/vlm/test_evs.py
index f95839e50..e2ee1c9e5 100644
--- a/test/registered/vlm/test_evs.py
+++ b/test/registered/vlm/test_evs.py
@@ -6,7 +6,7 @@ import pytest
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import run_doctests
 
-register_cuda_ci(est_time=11, stage="stage-b", runner_config="1-gpu-small")
+register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-small")
 register_amd_ci(est_time=20, suite="stage-b-test-1-gpu-small-amd")
 
 
diff --git a/test/registered/vlm/test_video_utils.py b/test/registered/vlm/test_video_utils.py
index 7587bb9ee..287845135 100644
--- a/test/registered/vlm/test_video_utils.py
+++ b/test/registered/vlm/test_video_utils.py
@@ -5,7 +5,7 @@ import pytest
 from sglang.srt.utils import sample_video_frames
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
+register_cpu_ci(est_time=7, suite="base-a-test-cpu")
 
 
 class DummyVideo:
diff --git a/test/registered/vlm/test_vision_chunked_prefill.py b/test/registered/vlm/test_vision_chunked_prefill.py
index 9f42d04e1..0f791decc 100644
--- a/test/registered/vlm/test_vision_chunked_prefill.py
+++ b/test/registered/vlm/test_vision_chunked_prefill.py
@@ -1,6 +1,6 @@
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=156, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=156, stage="base-b", runner_config="1-gpu-large")
 register_amd_ci(est_time=270, suite="stage-b-test-1-gpu-small-amd")
 """
 Usage:
diff --git a/test/registered/vlm/test_vision_openai_server_a.py b/test/registered/vlm/test_vision_openai_server_a.py
index 8611be701..68b1c79f2 100644
--- a/test/registered/vlm/test_vision_openai_server_a.py
+++ b/test/registered/vlm/test_vision_openai_server_a.py
@@ -19,7 +19,7 @@ from sglang.test.vlm_utils import (
     VideoOpenAITestMixin,
 )
 
-register_cuda_ci(est_time=780, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=780, stage="base-b", runner_config="1-gpu-large")
 
 
 class TestLlavaServer(ImageOpenAITestMixin):
diff --git a/test/registered/vlm/test_vlm_input_format.py b/test/registered/vlm/test_vlm_input_format.py
index 70c4cc4eb..94e387b10 100644
--- a/test/registered/vlm/test_vlm_input_format.py
+++ b/test/registered/vlm/test_vlm_input_format.py
@@ -38,7 +38,7 @@ from sglang.srt.parser.conversation import generate_chat_conv
 from sglang.srt.utils.common import is_cuda, is_xpu
 from sglang.srt.utils.hf_transformers_utils import _fix_added_tokens_encoding
 
-register_cuda_ci(est_time=747, stage="stage-b", runner_config="1-gpu-large")
+register_cuda_ci(est_time=747, stage="base-b", runner_config="1-gpu-large")
 
 IMAGE_MAN_IRONING_URL = "https://raw.githubusercontent.com/sgl-project/sgl-test-files/refs/heads/main/images/man_ironing_on_back_of_suv.png"
 IMAGE_SGL_LOGO_URL = "https://raw.githubusercontent.com/sgl-project/sgl-test-files/refs/heads/main/images/sgl_logo.png"
diff --git a/test/registered/vlm/test_vlm_tp4.py b/test/registered/vlm/test_vlm_tp4.py
index 22361383b..7aae595ee 100644
--- a/test/registered/vlm/test_vlm_tp4.py
+++ b/test/registered/vlm/test_vlm_tp4.py
@@ -15,7 +15,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=133, stage="stage-c", runner_config="4-gpu-h100")
+register_cuda_ci(est_time=133, stage="base-c", runner_config="4-gpu-h100")
 
 QWEN35_27B_MODEL = "Qwen/Qwen3.5-27B"
 MMMU_ACCURACY_THRESHOLD = 0.65
diff --git a/test/run_suite.py b/test/run_suite.py
index d1afb4e59..ceff2a2b7 100644
--- a/test/run_suite.py
+++ b/test/run_suite.py
@@ -22,9 +22,12 @@ HW_MAPPING = {
     "npu": HWBackend.NPU,
 }
 
-# Per-commit test suites (run on every PR)
+# Per-commit test suites (run on every PR).
+# Includes both base-a/b/c (always-on; pr-test.yml) and extra-a/b
+# (label-gated; pr-test-extra.yml). Tests are tagged per-commit regardless;
+# pr-test-extra.yml's `run-ci-extra` PR label decides whether extra-* dispatches.
 PER_COMMIT_SUITES = {
-    HWBackend.CPU: ["stage-a-test-cpu", "stage-b-test-cpu"],
+    HWBackend.CPU: ["base-a-test-cpu", "base-b-test-cpu"],
     HWBackend.AMD: [
         "stage-a-test-1-gpu-small-amd",
         "stage-b-test-1-gpu-small-amd",
@@ -39,24 +42,24 @@ PER_COMMIT_SUITES = {
         "stage-c-test-large-8-gpu-amd-mi35x",
     ],
     HWBackend.CUDA: [
-        "stage-a-test-1-gpu-small",
-        "stage-b-test-1-gpu-small",
-        "stage-b-test-1-gpu-large",
-        "stage-b-test-2-gpu-large",
-        "stage-b-test-4-gpu-b200",
-        "stage-b-kernel-unit-1-gpu-large",
-        "stage-b-kernel-unit-1-gpu-b200",
-        "stage-b-kernel-unit-8-gpu-h200",
-        "stage-b-kernel-benchmark-1-gpu-large",
-        "stage-c-test-4-gpu-h100",
-        "stage-c-test-4-gpu-b200",
-        "stage-c-test-4-gpu-gb200",
-        "stage-c-test-8-gpu-h20",
-        "stage-c-test-8-gpu-h200",
-        "stage-c-test-8-gpu-b200",
-        "stage-c-test-deepep-4-gpu-h100",
-        "stage-c-test-dsv4-4-gpu-b200",
-        "stage-c-test-dsv4-8-gpu-h200",
+        "base-a-test-1-gpu-small",
+        "base-b-test-1-gpu-small",
+        "base-b-test-1-gpu-large",
+        "base-b-test-2-gpu-large",
+        "base-b-test-4-gpu-b200",
+        "base-b-kernel-unit-1-gpu-large",
+        "base-b-kernel-unit-1-gpu-b200",
+        "base-b-kernel-unit-8-gpu-h200",
+        "base-b-kernel-benchmark-1-gpu-large",
+        "base-c-test-4-gpu-h100",
+        "base-c-test-4-gpu-b200",
+        "base-c-test-4-gpu-gb200",
+        "base-c-test-8-gpu-h20",
+        "base-c-test-8-gpu-h200",
+        "base-c-test-8-gpu-b200",
+        "base-c-test-deepep-4-gpu-h100",
+        "base-c-test-dsv4-4-gpu-b200",
+        "base-c-test-dsv4-8-gpu-h200",
         # extra-a / extra-b: label-gated PR opt-in suites in pr-test-extra.yml
         # (tests still tagged per-commit but skipped on default PR runs).
         "extra-a-test-1-gpu-small",
@@ -68,7 +71,7 @@ PER_COMMIT_SUITES = {
         "extra-b-test-deepep-8-gpu-h200",
     ],
     HWBackend.NPU: [
-        "stage-a-test-1-gpu-small",
+        "base-a-test-1-gpu-small",
         "stage-b-test-1-npu-a2",
         "stage-b-test-2-npu-a2",
         "stage-b-test-4-npu-a3",
diff --git a/test/srt/run_suite.py b/test/srt/run_suite.py
index 00af66144..61b41e422 100644
--- a/test/srt/run_suite.py
+++ b/test/srt/run_suite.py
@@ -9,7 +9,7 @@ from sglang.test.ci.ci_utils import TestFile, run_unittest_files
 # NOTE: please sort the test cases alphabetically by the test file name
 # NOTE: per-commit-4-gpu, per-commit-8-gpu-h200, per-commit-8-gpu-h20, per-commit-4-gpu-b200,
 # per-commit-4-gpu-gb200, per-commit-4-gpu-deepep, and per-commit-8-gpu-h200-deepep suites
-# have been migrated to stage-c suites in test/registered/ using the CI registry system.
+# have been migrated to base-c suites in test/registered/ using the CI registry system.
 suites = {
     # quantization_test suite migrated to test/registered/quant/
     # All CUDA tests migrated to test/registered/