diff --git a/docs_new/cookbook/autoregressive/Poolside/Laguna-M.1.mdx b/docs_new/cookbook/autoregressive/Poolside/Laguna-M.1.mdx
index ea643df4a..32976dcbe 100644
--- a/docs_new/cookbook/autoregressive/Poolside/Laguna-M.1.mdx
+++ b/docs_new/cookbook/autoregressive/Poolside/Laguna-M.1.mdx
@@ -1,7 +1,6 @@
---
title: Laguna-M.1
description: "Deploy poolside's Laguna-M.1 — a 225B-parameter Mixture-of-Experts model (23B active) for agentic coding — with SGLang on NVIDIA H200, B200, B300, GB200, and GB300, across BF16, FP8, and NVFP4."
-tag: NEW
---
## Deployment
diff --git a/docs_new/cookbook/autoregressive/Poolside/Laguna-S-2.1.mdx b/docs_new/cookbook/autoregressive/Poolside/Laguna-S-2.1.mdx
new file mode 100644
index 000000000..1d4f13b36
--- /dev/null
+++ b/docs_new/cookbook/autoregressive/Poolside/Laguna-S-2.1.mdx
@@ -0,0 +1,244 @@
+---
+title: Laguna-S-2.1
+description: "Deploy poolside's Laguna-S-2.1 — a 118B hybrid-SWA Mixture-of-Experts model (8B active) for agentic coding — with SGLang on NVIDIA H200, B300, and GB300 in BF16, FP8, NVFP4, and INT4."
+tag: NEW
+---
+
+## Deployment
+
+
+
+
+
+Laguna-S-2.1 uses the same `laguna` model architecture as [Laguna-XS-2.1](./Laguna-XS-2.1), which is fully supported in SGLang `main`. The model ships custom config code on the Hub, so `--trust-remote-code` is required (included in the launch commands).
+
+
+
+
+
+```bash Command
+pip install -U uv
+uv venv --python 3.12 && source .venv/bin/activate
+
+git clone https://github.com/sgl-project/sglang.git
+cd sglang
+uv pip install -e python
+```
+
+Then run the **Python** output of the command panel below in that environment.
+
+
+
+
+
+```bash Command
+docker pull lmsysorg/sglang:latest
+```
+
+For how to launch the image, see [Install → Method 3: Using Docker](../../../docs/get-started/install#method-3-using-docker). Substitute the inner `sglang serve ...` with what the command generator below produces.
+
+
+
+
+
+
+
+Pick your hardware + quantization + strategy to generate the launch command. The two serving strategies cover the common operating points:
+
+- **Low-latency** — DFlash speculative decoding with a matched draft model. Pick for chat and interactive agents.
+- **High-throughput** — plain serving. Best for batch workloads, where speculation's draft + rejection overhead costs more than it saves.
+
+On the 8-GPU HGX platforms (H200 / B300) all quantizations run `--tp 8`. The 4-GPU GB300 node runs `--tp 4` throughout. NVFP4 is Blackwell-only (B300 / GB300 only).
+
+import { Deployment } from "/src/snippets/_deployment.jsx";
+import { config } from "/src/snippets/configs/poolside/laguna-s21.jsx";
+import { benchmarks } from "/src/snippets/configs/poolside/laguna-s21-benchmarks.jsx";
+
+
+
+## Playground
+
+The Playground is where you experiment with **SGLang features beyond the verified matrix**. The Deploy panel above only emits combinations that have been signed off; the Playground lets you turn on additional knobs (TP degree, parsers) on top of whichever cell the Deploy panel is currently showing.
+
+import { Playground } from "/src/snippets/_playground.jsx";
+
+
+
+## 1. Model Introduction
+
+[Laguna-S-2.1](https://huggingface.co/poolside/Laguna-S-2.1) is an open-weight **118B-parameter** hybrid sliding-window-attention MoE model (**~8B active per token**) from [poolside](https://poolside.ai), built for agentic coding and long-horizon software engineering. It sits between [Laguna XS 2.1](./Laguna-XS-2.1) (33B/3B active) and Laguna M.1 (222B/23B active) in the Laguna family.
+
+**Key Features:**
+
+- **Sparse MoE**: 48 layers, 256 routed experts, top-10 routing, plus 1 shared expert.
+- **Hybrid attention**: 36 sliding-window layers (window 512) interleaved with 12 full-attention layers (1:3 global-to-SWA ratio); 8 KV heads, head dim 128; per-head sigmoid output gating with per-layer-type rotary scales.
+- **Long context**: 1,048,576 tokens.
+- **DFlash drafts**: matched draft models ship per quantization for low-latency serving.
+- **Hybrid reasoning**: `…` toggled per request via `chat_template_kwargs={"enable_thinking": …}`.
+
+**Available quantizations:**
+
+
+
+
+
+
+
+
+
+ | Precision |
+ Target model |
+ Draft model |
+
+
+
+
+ | BF16 |
+ [`poolside/Laguna-S-2.1`](https://huggingface.co/poolside/Laguna-S-2.1) |
+ [`poolside/Laguna-S-2.1-DFlash`](https://huggingface.co/poolside/Laguna-S-2.1-DFlash) |
+
+
+ | FP8 |
+ [`poolside/Laguna-S-2.1-FP8`](https://huggingface.co/poolside/Laguna-S-2.1-FP8) |
+ [`poolside/Laguna-S-2.1-DFlash-FP8`](https://huggingface.co/poolside/Laguna-S-2.1-DFlash-FP8) |
+
+
+ | NVFP4 |
+ [`poolside/Laguna-S-2.1-NVFP4`](https://huggingface.co/poolside/Laguna-S-2.1-NVFP4) |
+ [`poolside/Laguna-S-2.1-DFlash-NVFP4`](https://huggingface.co/poolside/Laguna-S-2.1-DFlash-NVFP4) |
+
+
+ | INT4 |
+ [`poolside/Laguna-S-2.1-INT4`](https://huggingface.co/poolside/Laguna-S-2.1-INT4) |
+ [`poolside/Laguna-S-2.1-DFlash-INT4`](https://huggingface.co/poolside/Laguna-S-2.1-DFlash-INT4) |
+
+
+
+
+The drafts are small BF16 models, each *calibrated against its quantized target* — always pair a target with its matched draft (mixing precisions degrades accept-length).
+
+**License:** [OpenMDW-1.1](https://openmdw.ai/)
+
+**Resources:** [Hugging Face](https://huggingface.co/poolside/Laguna-S-2.1) · [Technical report](https://poolside.ai/assets/laguna/laguna-m1-xs2-technical-report.pdf) · [API platform](https://platform.poolside.ai)
+
+## 2. Configuration Tips
+
+**Attention backend**
+
+Leave `--attention-backend` unset for High-throughput cells — auto-select is correct (`fa3` on Hopper, `trtllm_mha` on Blackwell). With DFlash active, auto-select instead falls back to `flashinfer`, which breaks this hybrid-SWA model at `tp ≥ 4` on Blackwell (reproduced on Laguna-XS-2.1, greedy GSM8K 76% → 28%), so the Low-latency commands pin the target backend explicitly. Leave `--speculative-draft-attention-backend` unset. Other attention backend choices have not been fully validated on Laguna; keep the default.
+
+**BF16 memory on H200**
+
+BF16 on H200 leaves less headroom for CUDA-graph capture and NCCL allocations than FP8/INT4. The High-throughput BF16 command carries `--mem-fraction-static 0.80`. FP8, INT4, and all B300/GB300 cells use the default heuristic.
+
+**FP8 shared expert**
+
+`SGLANG_SHARED_EXPERT_TP1=1` is required for FP8 cells on **all hardware** — confirmed on both H200 (TP=8) and GB300 (TP=4). The FP8 checkpoint block-quantizes the shared expert (128×128 scales), which cannot TP-shard cleanly at either TP degree on S-2.1. This env var replicates the shared expert instead of sharding it. INT4 keeps the shared expert in BF16 (no flag needed); BF16 is unquantized. Note: this differs from Laguna-XS-2.1 where TP=4 does not require the flag — the constraint is architecture-specific.
+
+**FP8 and NVFP4 DFlash drafts**
+
+Fixed upstream on 2026-07-21: all DFlash draft configs now use a flat top-level `rope_theta` (the `rope_parameters` block was removed). If a server crashes at draft-model load with `KeyError: 'rope_theta'`, you are serving a draft checkpoint cached before 2026-07-21 — re-download it (e.g. `hf download poolside/Laguna-S-2.1-DFlash-FP8`) to pick up the corrected config.
+
+**DFlash memory**
+
+Low-latency cells carry `--mem-fraction-static 0.7` (sufficient even for BF16 on H200). Dense cells use the default heuristic (except BF16 on H200 — see above).
+
+**BF16 reasoning length**
+
+BF16 reasons approximately 2× longer than FP8/INT4 on AIME25 (median 34.8 k vs 16.9 k tokens), consistently truncating at `max_tokens=64000`. FP8/INT4 truncate at ≈ 2%. For a valid BF16 AIME25 score, serve with `max_tokens ≥ 131072` (the model supports a 1 M context window).
+
+**Chat template**
+
+On transformers ≥ 5.10 the standalone `chat_template.jinja` auto-loads — no flag needed (the server logs `Auto-detected template features: reasoning_parser=poolside_v1, ...`). On older transformers (≤ ~5.8) pass `--chat-template /chat_template.jinja` explicitly.
+
+**Thinking**
+
+Off by default; opt in per request with `extra_body={"chat_template_kwargs": {"enable_thinking": True}}`. The template gates on `enable_thinking` — the generic `thinking` key is ignored.
+
+**Served model id**
+
+The server registers the model under whatever you pass to `--model-path`; a client's `model` field must match it (`poolside/Laguna-S-2.1`, or the `-FP8` / `-NVFP4` / `-INT4` id).
+
+## 3. Advanced Usage
+
+### 3.1 DFlash Speculative Decoding
+
+DFlash is a block-wise speculative decoder: the draft proposes a block of tokens and the target verifies the whole block in one forward pass — output quality is the target's by construction. The speedup lever is **accept-length**, the number of draft tokens surviving verification per target step.
+
+Best for interactive / few-stream serving. Under batch-saturated load prefer High-throughput: once the GPU is compute-bound, draft + rejected-token overhead costs aggregate throughput. The generated commands always pair the draft calibrated for the selected target precision.
+
+### 3.2 Reasoning
+
+Launch with `--reasoning-parser poolside_v1` (baked into every generated command). Reasoning is opt-in via `enable_thinking=True`; the `` trace lands in `message.reasoning_content`, separate from the final answer in `message.content`.
+
+
+
+```python Example
+from openai import OpenAI
+
+client = OpenAI(base_url="http://localhost:30000/v1", api_key="EMPTY")
+
+response = client.chat.completions.create(
+ model="poolside/Laguna-S-2.1",
+ messages=[{"role": "user", "content": "What is 15% of 240? Explain briefly."}],
+ max_tokens=4096,
+ extra_body={"chat_template_kwargs": {"enable_thinking": True}},
+)
+
+message = response.choices[0].message
+print("=============== Reasoning ===============")
+print(message.reasoning_content)
+print("=============== Answer ==================")
+print(message.content)
+```
+
+
+
+
+Give generous `max_tokens` when thinking is enabled — hard problems regularly reason
+for thousands of tokens. Keep thinking off for short-form tasks.
+
+
+### 3.3 Tool Calling
+
+Launch with `--tool-call-parser poolside_v1` (baked into every generated command). The parser converts Laguna's `` output into the standard OpenAI `tool_calls` structure. Tool calling works with reasoning off (the default).
+
+
+
+```python Example
+from openai import OpenAI
+
+client = OpenAI(base_url="http://localhost:30000/v1", api_key="EMPTY")
+
+tools = [
+ {
+ "type": "function",
+ "function": {
+ "name": "get_weather",
+ "description": "Get the current weather for a location",
+ "parameters": {
+ "type": "object",
+ "properties": {
+ "location": {"type": "string", "description": "The city name"},
+ "unit": {"type": "string", "enum": ["celsius", "fahrenheit"]},
+ },
+ "required": ["location"],
+ },
+ },
+ }
+]
+
+response = client.chat.completions.create(
+ model="poolside/Laguna-S-2.1",
+ messages=[{"role": "user", "content": "What's the weather in Beijing?"}],
+ tools=tools,
+)
+
+message = response.choices[0].message
+if message.tool_calls:
+ for call in message.tool_calls:
+ print(f"Tool: {call.function.name}")
+ print(f"Args: {call.function.arguments}")
+```
+
+
diff --git a/docs_new/cookbook/autoregressive/Poolside/Laguna-XS-2.1.mdx b/docs_new/cookbook/autoregressive/Poolside/Laguna-XS-2.1.mdx
index 5423eeda2..7d0ff41fa 100644
--- a/docs_new/cookbook/autoregressive/Poolside/Laguna-XS-2.1.mdx
+++ b/docs_new/cookbook/autoregressive/Poolside/Laguna-XS-2.1.mdx
@@ -1,7 +1,6 @@
---
title: Laguna-XS-2.1
description: "Deploy poolside's Laguna-XS-2.1 — a 33B hybrid-SWA Mixture-of-Experts model (3B active) for agentic coding — with SGLang on NVIDIA H200, B300, and GB300 in BF16, FP8, NVFP4, and INT4."
-tag: NEW
---
## Deployment
diff --git a/docs_new/cookbook/autoregressive/intro.mdx b/docs_new/cookbook/autoregressive/intro.mdx
index 5dac02921..3731e8e3f 100644
--- a/docs_new/cookbook/autoregressive/intro.mdx
+++ b/docs_new/cookbook/autoregressive/intro.mdx
@@ -160,7 +160,7 @@ metatags:
diff --git a/docs_new/docs.json b/docs_new/docs.json
index 1c2d81bcf..c01f7bb5e 100644
--- a/docs_new/docs.json
+++ b/docs_new/docs.json
@@ -1172,6 +1172,7 @@
{
"group": "Poolside",
"pages": [
+ "cookbook/autoregressive/Poolside/Laguna-S-2.1",
"cookbook/autoregressive/Poolside/Laguna-XS-2.1",
"cookbook/autoregressive/Poolside/Laguna-M.1",
"cookbook/autoregressive/Poolside/Laguna-XS.2"
diff --git a/docs_new/src/snippets/configs/poolside/laguna-s21-benchmarks.jsx b/docs_new/src/snippets/configs/poolside/laguna-s21-benchmarks.jsx
new file mode 100644
index 000000000..73fdc3589
--- /dev/null
+++ b/docs_new/src/snippets/configs/poolside/laguna-s21-benchmarks.jsx
@@ -0,0 +1,218 @@
+export const benchmarks = [
+
+ // ── H200 (8×H200, tp 8, sglang 0.5.15.post1, random ISL=8192/OSL=1024) ──
+ // tokens_per_sec_per_gpu = (input+output) tok/s/GPU = output_tok_s / 8 * (8192+1024)/1024
+ // TTFT/TPOT are mean values from bench_serving.
+
+ {
+ match: { hw: "h200", variant: "default", quant: "bf16", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ speed: [
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 1024 },
+ ttft_ms: 23742, tpot_ms: 256.8, tokens_per_sec_per_gpu: 5264 },
+ ],
+ // BF16 reasons ~2× longer than FP8/INT4 (median 34.8k vs 16.9k tokens);
+ // truncation at max_tokens=64000 invalidates the result. Needs max_tokens ≥ 131072.
+ accuracy: { gsm8k_pct: 93.18, aime25_pct: null },
+ },
+ {
+ match: { hw: "h200", variant: "default", quant: "bf16", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ speed: [
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 1 },
+ ttft_ms: 102.3, tpot_ms: 3.48, tokens_per_sec_per_gpu: 305 },
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 16 },
+ ttft_ms: 97.7, tpot_ms: 6.45, tokens_per_sec_per_gpu: 2094 },
+ ],
+ accuracy: { gsm8k_pct: 93.33, aime25_pct: null },
+ },
+ {
+ match: { hw: "h200", variant: "default", quant: "fp8", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ speed: [
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 1024 },
+ ttft_ms: 16199, tpot_ms: 293.2, tokens_per_sec_per_gpu: 5175 },
+ ],
+ accuracy: { gsm8k_pct: 94.24, aime25_pct: 0.654 },
+ },
+ {
+ match: { hw: "h200", variant: "default", quant: "fp8", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ speed: [
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 1 },
+ ttft_ms: 97.7, tpot_ms: 4.43, tokens_per_sec_per_gpu: 242 },
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 16 },
+ ttft_ms: 99.4, tpot_ms: 7.16, tokens_per_sec_per_gpu: 1956 },
+ ],
+ accuracy: { gsm8k_pct: 94.47, aime25_pct: null },
+ },
+ {
+ match: { hw: "h200", variant: "default", quant: "int4", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ speed: [
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 1024 },
+ ttft_ms: 19886, tpot_ms: 318.5, tokens_per_sec_per_gpu: 5055 },
+ ],
+ accuracy: { gsm8k_pct: 95.00, aime25_pct: 0.690 },
+ },
+ {
+ match: { hw: "h200", variant: "default", quant: "int4", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ speed: [
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 1 },
+ ttft_ms: 113.1, tpot_ms: 3.83, tokens_per_sec_per_gpu: 276 },
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 16 },
+ ttft_ms: 94.1, tpot_ms: 6.40, tokens_per_sec_per_gpu: 2125 },
+ ],
+ accuracy: { gsm8k_pct: 94.24, aime25_pct: null },
+ },
+
+ // ── B300 (8×B300, tp 8, sglang 0.5.15.post1) ──
+ // AIME25 (high-throughput cells): sgl-eval run aime25, 30 problems × 16 repeats,
+ // temp 1.0, top_p 0.95, max_tokens 64000 (BF16: 131072), 128 threads, thinking ON
+ // via enable_thinking-patched sgl-eval. 2026-07-21.
+
+ {
+ match: { hw: "b300", variant: "default", quant: "bf16", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ accuracy: { gsm8k_pct: 93.71, aime25_pct: null },
+ },
+ {
+ match: { hw: "b300", variant: "default", quant: "bf16", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ accuracy: { gsm8k_pct: 93.63, aime25_pct: null },
+ },
+ {
+ match: { hw: "b300", variant: "default", quant: "fp8", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ accuracy: { gsm8k_pct: 94.39, aime25_pct: 66.88 },
+ },
+ {
+ match: { hw: "b300", variant: "default", quant: "fp8", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ accuracy: { gsm8k_pct: 94.69, aime25_pct: null },
+ },
+ {
+ match: { hw: "b300", variant: "default", quant: "nvfp4", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ accuracy: { gsm8k_pct: 94.54, aime25_pct: 66.46 },
+ },
+ {
+ match: { hw: "b300", variant: "default", quant: "nvfp4", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ accuracy: { gsm8k_pct: 95.30, aime25_pct: null },
+ },
+ {
+ match: { hw: "b300", variant: "default", quant: "int4", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ accuracy: { gsm8k_pct: 94.62, aime25_pct: 68.33 },
+ },
+ {
+ match: { hw: "b300", variant: "default", quant: "int4", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ accuracy: { gsm8k_pct: 94.69, aime25_pct: null },
+ },
+
+ // ── GB300 (4×GB300, tp 4, sglang 0.5.15.post1, random ISL=8192/OSL=1024) ──
+ // tokens_per_sec_per_gpu = output_tok_s / 4 * (8192+1024)/1024
+ // TTFT/TPOT are median values from bench_serving.
+
+ {
+ match: { hw: "gb300", variant: "default", quant: "bf16", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ accuracy: { gsm8k_pct: 93.33, aime25_pct: null },
+ },
+ {
+ match: { hw: "gb300", variant: "default", quant: "bf16", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ speed: [
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 1 },
+ ttft_ms: 114, tpot_ms: 4.2, tokens_per_sec_per_gpu: 562 },
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 16 },
+ ttft_ms: 141, tpot_ms: 8.4, tokens_per_sec_per_gpu: 3413 },
+ ],
+ accuracy: { gsm8k_pct: 93.33, aime25_pct: null },
+ },
+ {
+ match: { hw: "gb300", variant: "default", quant: "fp8", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ speed: [
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 1024 },
+ ttft_ms: 10062, tpot_ms: 216, tokens_per_sec_per_gpu: 9536 },
+ ],
+ accuracy: { gsm8k_pct: 94.31, aime25_pct: null },
+ },
+ {
+ match: { hw: "gb300", variant: "default", quant: "fp8", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ speed: [
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 1 },
+ ttft_ms: 109, tpot_ms: 6.1, tokens_per_sec_per_gpu: 369 },
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 16 },
+ ttft_ms: 135, tpot_ms: 10.7, tokens_per_sec_per_gpu: 2578 },
+ ],
+ accuracy: { gsm8k_pct: 94.54, aime25_pct: null },
+ },
+ {
+ match: { hw: "gb300", variant: "default", quant: "nvfp4", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ speed: [
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 1024 },
+ ttft_ms: 7186, tpot_ms: 211, tokens_per_sec_per_gpu: 9658 },
+ ],
+ accuracy: { gsm8k_pct: 94.47, aime25_pct: null },
+ },
+ {
+ match: { hw: "gb300", variant: "default", quant: "nvfp4", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ speed: [
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 1 },
+ ttft_ms: 115, tpot_ms: 11.2, tokens_per_sec_per_gpu: 204 },
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 16 },
+ ttft_ms: 125, tpot_ms: 18.8, tokens_per_sec_per_gpu: 1549 },
+ ],
+ accuracy: { gsm8k_pct: 94.77, aime25_pct: null },
+ },
+ {
+ match: { hw: "gb300", variant: "default", quant: "int4", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ speed: [
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 1024 },
+ ttft_ms: 10428, tpot_ms: 216, tokens_per_sec_per_gpu: 9184 },
+ ],
+ accuracy: { gsm8k_pct: 94.69, aime25_pct: null },
+ },
+ {
+ match: { hw: "gb300", variant: "default", quant: "int4", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ sglang_version: "0.5.15.post1",
+ speed: [
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 1 },
+ ttft_ms: 134, tpot_ms: 5.4, tokens_per_sec_per_gpu: 410 },
+ { workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 16 },
+ ttft_ms: 127, tpot_ms: 10.1, tokens_per_sec_per_gpu: 2880 },
+ ],
+ accuracy: { gsm8k_pct: 95.00, aime25_pct: null },
+ },
+];
diff --git a/docs_new/src/snippets/configs/poolside/laguna-s21.jsx b/docs_new/src/snippets/configs/poolside/laguna-s21.jsx
new file mode 100644
index 000000000..cdf9ac360
--- /dev/null
+++ b/docs_new/src/snippets/configs/poolside/laguna-s21.jsx
@@ -0,0 +1,506 @@
+// Laguna-S-2.1 (poolside) — 118B MoE (8B active), 1M context, laguna arch (SGLang main).
+// --trust-remote-code required (custom config code on Hub).
+//
+// Attention backend: leave unset for High-Throughput (auto-selects fa3/trtllm_mha).
+// With DFlash active, auto falls back to flashinfer which breaks hybrid-SWA at tp≥4
+// on Blackwell — Low-Latency cells pin the target backend explicitly.
+// Never use --attention-backend triton on Laguna (broken SWA handling).
+//
+// BF16 on H200 HT: --mem-fraction-static 0.80 required — BF16 leaves less headroom
+// for CUDA-graph capture and NCCL allocs than FP8/INT4 on 141 GB/GPU.
+// B300/GB300 (288 GB/GPU) unaffected.
+//
+// SGLANG_SHARED_EXPERT_TP1=1 (FP8 cells, all hardware): FP8 block-quantizes the shared
+// expert; required at both TP=4 (GB300) and TP=8 (H200/B300). INT4 shared expert stays
+// bf16 — no flag needed. Differs from Laguna-XS-2.1 where TP=4 works without this flag.
+//
+// NVFP4 is Blackwell-only → no h200×nvfp4 cells.
+// DFlash cells carry --mem-fraction-static 0.7.
+
+export const config = {
+ modelName: "Laguna-S-2.1",
+
+ supportedHardware: ["h200", "b300", "gb300"],
+
+ variants: [
+ { id: "default", label: "Default" },
+ ],
+
+ quantizations: [
+ { id: "bf16", label: "BF16" },
+ { id: "fp8", label: "FP8" },
+ { id: "nvfp4", label: "NVFP4" },
+ { id: "int4", label: "INT4" },
+ ],
+
+ strategies: [
+ { id: "low-latency", label: "Low-latency" },
+ { id: "high-throughput", label: "High-throughput" },
+ ],
+
+ nodesOptions: [
+ { id: "single", label: "Single Node" },
+ ],
+
+ modelNames: {
+ "default|bf16": "poolside/Laguna-S-2.1",
+ "default|fp8": "poolside/Laguna-S-2.1-FP8",
+ "default|nvfp4": "poolside/Laguna-S-2.1-NVFP4",
+ "default|int4": "poolside/Laguna-S-2.1-INT4",
+ },
+
+ placeholders: {
+ HOST_IP: { target: "command", label: "Bind host", default: "0.0.0.0" },
+ PORT: { target: "command", label: "Bind port", default: "30000" },
+ HF_TOKEN: { target: "command", label: "HF token (Docker)", default: "" },
+ CURL_HOST: { target: "curl", label: "Server host", default: "localhost" },
+ CURL_PORT: { target: "curl", label: "Server port", default: "30000" },
+ },
+
+ curl: `curl http://{{CURL_HOST}}:{{CURL_PORT}}/v1/chat/completions \\
+-H 'Content-Type: application/json' \\
+-d '{ "model": "{{MODEL_NAME}}", "messages": [{"role":"user","content":"Hello"}] }'`,
+
+ benchmarkCommands: {
+ speed:
+`python3 -m sglang.bench_serving \\
+ --backend sglang \\
+ --host {{CURL_HOST}} --port {{CURL_PORT}} \\
+ --model {{MODEL_NAME}} \\
+ --dataset-name {{DATASET}} \\
+ --random-input-len {{ISL}} --random-output-len {{OSL}} \\
+ --num-prompts {{NUM_PROMPTS}} --max-concurrency {{MAX_CONCURRENCY}}`,
+ accuracy: {
+ gsm8k_pct:
+`# pip install git+https://github.com/sgl-project/sgl-eval
+sgl-eval run gsm8k \\
+ --base-url http://{{CURL_HOST}}:{{CURL_PORT}}/v1 \\
+ --num-threads 128`,
+ // Laguna's template gates on enable_thinking, not the generic 'thinking' key.
+ // Serve with a copy of the model's chat template whose enable_thinking default
+ // is flipped to true. For BF16: use --max-tokens 131072 (see Configuration Tips).
+ aime25_pct:
+`# pip install git+https://github.com/sgl-project/sgl-eval
+# Serve with an enable_thinking=true chat template (see Configuration Tips: Thinking).
+# For BF16: use --max-tokens 131072 (see Configuration Tips: BF16 reasoning length).
+sgl-eval run aime25 \\
+ --base-url http://{{CURL_HOST}}:{{CURL_PORT}}/v1 \\
+ --n-repeats 16 --max-tokens 64000 \\
+ --temperature 1.0 --top-p 0.95 --thinking \\
+ --num-threads 128`,
+ },
+ numPromptsByConc: { 1: 8, 16: 32, 64: 128, 128: 256, 256: 512, 1024: 2048, 4096: 4096 },
+ },
+
+ defaultAccuracy: {
+ default: { gsm8k_pct: null, aime25_pct: null },
+ },
+
+ accuracyLabels: [
+ ["gsm8k_pct", "GSM8K", "%"],
+ ["aime25_pct", "AIME25", "%"],
+ ],
+
+ dockerImages: {
+ h200: "lmsysorg/sglang:latest",
+ b300: "lmsysorg/sglang:dev",
+ gb300: "lmsysorg/sglang:dev",
+ },
+
+ github: {
+ cookbookModel: "poolside/Laguna-S-2.1",
+ },
+
+ playgroundFeatures: {
+ attention: {
+ knobs: [
+ { id: "tp", label: "TP", values: [null, 1, 2, 4, 8] },
+ ],
+ },
+ parsers: {
+ items: [
+ { id: "reasoning", label: "Reasoning Parser", flag: "--reasoning-parser poolside_v1" },
+ { id: "toolCall", label: "Tool Call Parser", flag: "--tool-call-parser poolside_v1" },
+ ],
+ },
+ },
+
+ cells: [
+
+ // ══════════════ B300 FP8 low-latency — default (cells[0]) ══════════════
+
+ {
+ match: { hw: "b300", variant: "default", quant: "fp8", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ env: ["SGLANG_SHARED_EXPERT_TP1=1"],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 8",
+ "--attention-backend trtllm_mha",
+ "--speculative-algorithm DFLASH",
+ "--speculative-draft-model-path poolside/Laguna-S-2.1-DFlash-FP8",
+ "--page-size 1",
+ "--mem-fraction-static 0.7",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+
+ // ══════════════ H200 (8-GPU HGX, tp 8) ══════════════
+
+ {
+ match: { hw: "h200", variant: "default", quant: "bf16", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ env: [],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 8",
+ "--mem-fraction-static 0.80",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "h200", variant: "default", quant: "bf16", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ env: [],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 8",
+ "--attention-backend fa3",
+ "--speculative-algorithm DFLASH",
+ "--speculative-draft-model-path poolside/Laguna-S-2.1-DFlash",
+ "--page-size 1",
+ "--mem-fraction-static 0.7",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "h200", variant: "default", quant: "fp8", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ env: ["SGLANG_SHARED_EXPERT_TP1=1"],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 8",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "h200", variant: "default", quant: "fp8", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ env: ["SGLANG_SHARED_EXPERT_TP1=1"],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 8",
+ "--attention-backend fa3",
+ "--speculative-algorithm DFLASH",
+ "--speculative-draft-model-path poolside/Laguna-S-2.1-DFlash-FP8",
+ "--page-size 1",
+ "--mem-fraction-static 0.7",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "h200", variant: "default", quant: "int4", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ env: [],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 8",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "h200", variant: "default", quant: "int4", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ env: [],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 8",
+ "--attention-backend fa3",
+ "--speculative-algorithm DFLASH",
+ "--speculative-draft-model-path poolside/Laguna-S-2.1-DFlash-INT4",
+ "--page-size 1",
+ "--mem-fraction-static 0.7",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+
+ // ══════════════ B300 (8-GPU HGX, tp 8) ══════════════
+
+ {
+ match: { hw: "b300", variant: "default", quant: "bf16", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ env: [],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 8",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "b300", variant: "default", quant: "bf16", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ env: [],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 8",
+ "--attention-backend trtllm_mha",
+ "--speculative-algorithm DFLASH",
+ "--speculative-draft-model-path poolside/Laguna-S-2.1-DFlash",
+ "--page-size 1",
+ "--mem-fraction-static 0.7",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "b300", variant: "default", quant: "fp8", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ env: ["SGLANG_SHARED_EXPERT_TP1=1"],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 8",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "b300", variant: "default", quant: "nvfp4", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ env: [],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 8",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "b300", variant: "default", quant: "nvfp4", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ env: [],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 8",
+ "--attention-backend trtllm_mha",
+ "--speculative-algorithm DFLASH",
+ "--speculative-draft-model-path poolside/Laguna-S-2.1-DFlash-NVFP4",
+ "--page-size 1",
+ "--mem-fraction-static 0.7",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "b300", variant: "default", quant: "int4", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ env: [],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 8",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "b300", variant: "default", quant: "int4", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ env: [],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 8",
+ "--attention-backend trtllm_mha",
+ "--speculative-algorithm DFLASH",
+ "--speculative-draft-model-path poolside/Laguna-S-2.1-DFlash-INT4",
+ "--page-size 1",
+ "--mem-fraction-static 0.7",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+
+ // ══════════════ GB300 (4-GPU single node, tp 4) ══════════════
+
+ {
+ match: { hw: "gb300", variant: "default", quant: "bf16", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ env: [],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 4",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "gb300", variant: "default", quant: "bf16", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ env: [],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 4",
+ "--attention-backend trtllm_mha",
+ "--speculative-algorithm DFLASH",
+ "--speculative-draft-model-path poolside/Laguna-S-2.1-DFlash",
+ "--page-size 1",
+ "--mem-fraction-static 0.7",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "gb300", variant: "default", quant: "fp8", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ env: ["SGLANG_SHARED_EXPERT_TP1=1"],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 4",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "gb300", variant: "default", quant: "fp8", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ env: ["SGLANG_SHARED_EXPERT_TP1=1"],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 4",
+ "--attention-backend trtllm_mha",
+ "--speculative-algorithm DFLASH",
+ "--speculative-draft-model-path poolside/Laguna-S-2.1-DFlash-FP8",
+ "--page-size 1",
+ "--mem-fraction-static 0.7",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "gb300", variant: "default", quant: "nvfp4", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ env: [],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 4",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "gb300", variant: "default", quant: "nvfp4", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ env: [],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 4",
+ "--attention-backend trtllm_mha",
+ "--speculative-algorithm DFLASH",
+ "--speculative-draft-model-path poolside/Laguna-S-2.1-DFlash-NVFP4",
+ "--page-size 1",
+ "--mem-fraction-static 0.7",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "gb300", variant: "default", quant: "int4", strategy: "high-throughput", nodes: "single" },
+ verified: true,
+ env: [],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 4",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ {
+ match: { hw: "gb300", variant: "default", quant: "int4", strategy: "low-latency", nodes: "single" },
+ verified: true,
+ env: [],
+ flags: [
+ "--model-path {{MODEL_NAME}}",
+ "--trust-remote-code",
+ "--reasoning-parser poolside_v1",
+ "--tool-call-parser poolside_v1",
+ "--tp 4",
+ "--attention-backend trtllm_mha",
+ "--speculative-algorithm DFLASH",
+ "--speculative-draft-model-path poolside/Laguna-S-2.1-DFlash-INT4",
+ "--page-size 1",
+ "--mem-fraction-static 0.7",
+ "--host {{HOST_IP}}",
+ "--port {{PORT}}",
+ ],
+ },
+ ],
+};