diff --git a/.github/workflows/pr-test.yml b/.github/workflows/pr-test.yml
index 1887be7a3..0f183b20c 100644
--- a/.github/workflows/pr-test.yml
+++ b/.github/workflows/pr-test.yml
@@ -388,7 +388,6 @@ jobs:
       caller_inputs: ${{ toJson(inputs) }}
       partitions: ${{ needs.check-changes.outputs.partitions }}
       run_timeout_minutes: '40'
-      extra_pytest_path: 'test/registered/jit/test_flash_attention_4.py'
     secrets: inherit
 
   call-multimodal-gen-tests:
diff --git a/test/registered/jit/test_flash_attention_4.py b/test/registered/jit/test_flash_attention_4.py
index 95bb978af..a3d45b6bf 100644
--- a/test/registered/jit/test_flash_attention_4.py
+++ b/test/registered/jit/test_flash_attention_4.py
@@ -15,6 +15,7 @@ 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="base-b-kernel-unit-1-gpu-large")
+register_cuda_ci(est_time=120, suite="base-b-kernel-unit-1-gpu-b200")
 register_cuda_ci(est_time=900, suite="nightly-kernel-1-gpu", nightly=True)
 
 # Skip this test on Hopper machine