diff --git a/.github/workflows/pr-test.yml b/.github/workflows/pr-test.yml
index 74446bd1e..d621053da 100644
--- a/.github/workflows/pr-test.yml
+++ b/.github/workflows/pr-test.yml
@@ -609,6 +609,48 @@ jobs:
           cd test/srt
           python3 run_suite.py --suite per-commit-1-gpu --auto-partition-id ${{ matrix.part }} --auto-partition-size 15
 
+
+  stage-b-test-4-gpu-b200:
+    needs: [check-changes, call-gate, unit-test-backend-1-gpu]
+    if: |
+      always() &&
+      (
+        (inputs.target_stage == 'stage-b-test-4-gpu-b200') ||
+        (
+          !inputs.target_stage &&
+          (github.event_name == 'schedule' || (!failure() && !cancelled())) &&
+          ((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
+        )
+      )
+    runs-on: ${{ needs.check-changes.outputs.b200_runner }}
+    env:
+      RUNNER_LABELS: ${{ needs.check-changes.outputs.b200_runner }}
+    strategy:
+      fail-fast: false
+
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+
+      - name: Download artifacts
+        if: needs.check-changes.outputs.sgl_kernel == 'true'
+        uses: actions/download-artifact@v6
+        with:
+          path: sgl-kernel/dist/
+          merge-multiple: true
+          pattern: wheel-python3.10-cuda12.9
+
+      - name: Install dependencies
+        run: |
+          CUSTOM_BUILD_SGL_KERNEL=${{needs.check-changes.outputs.sgl_kernel}} IS_BLACKWELL=1 bash scripts/ci/ci_install_dependency.sh
+
+      - name: Run test
+        timeout-minutes: 30
+        run: |
+          cd test/srt
+          IS_BLACKWELL=1 python3 run_suite.py --suite per-commit-4-gpu-b200-stage-b --timeout-per-file 1800
+
+
   unit-test-backend-2-gpu:
     needs: [check-changes, call-gate, unit-test-backend-1-gpu]
     if: |
@@ -651,7 +693,7 @@ jobs:
           python3 run_suite.py --suite per-commit-2-gpu --auto-partition-id ${{ matrix.part }} --auto-partition-size 2
 
   unit-test-backend-4-gpu:
-    needs: [check-changes, call-gate, unit-test-backend-2-gpu]
+    needs: [check-changes, call-gate, unit-test-backend-2-gpu, stage-b-test-4-gpu-b200]
     if: |
       always() &&
       (
@@ -692,7 +734,7 @@ jobs:
           python3 run_suite.py --suite per-commit-4-gpu --auto-partition-id ${{ matrix.part }} --auto-partition-size 3
 
   unit-test-backend-8-gpu-h200:
-    needs: [check-changes, call-gate, unit-test-backend-2-gpu]
+    needs: [check-changes, call-gate, unit-test-backend-2-gpu, stage-b-test-4-gpu-b200]
     if: |
       always() &&
       (
@@ -739,7 +781,7 @@ jobs:
           python3 run_suite.py --suite per-commit-8-gpu-h200 --auto-partition-id ${{ matrix.part }} --auto-partition-size 3
 
   unit-test-backend-8-gpu-h20:
-    needs: [check-changes, call-gate, unit-test-backend-2-gpu]
+    needs: [check-changes, call-gate, unit-test-backend-2-gpu, stage-b-test-4-gpu-b200]
     if: |
       always() &&
       (
@@ -966,7 +1008,7 @@ jobs:
           python3 -m unittest test_bench_serving.TestBenchServing.test_embeddings_api_batch_scaling
 
   performance-test-2-gpu:
-    needs: [check-changes, call-gate, unit-test-backend-2-gpu]
+    needs: [check-changes, call-gate, unit-test-backend-2-gpu, stage-b-test-4-gpu-b200]
     if: |
       always() &&
       (
@@ -1113,7 +1155,7 @@ jobs:
           python3 test_moe_eval_accuracy_large.py
 
   unit-test-deepep-4-gpu:
-    needs: [check-changes, call-gate, unit-test-backend-2-gpu]
+    needs: [check-changes, call-gate, unit-test-backend-2-gpu, stage-b-test-4-gpu-b200]
     if: |
       always() &&
       (
@@ -1150,7 +1192,7 @@ jobs:
           python3 run_suite.py --suite per-commit-4-gpu-deepep
 
   unit-test-deepep-8-gpu:
-    needs: [check-changes, call-gate, unit-test-backend-2-gpu]
+    needs: [check-changes, call-gate, unit-test-backend-2-gpu, stage-b-test-4-gpu-b200]
     if: |
       always() &&
       (
@@ -1187,7 +1229,7 @@ jobs:
           python3 run_suite.py --suite per-commit-8-gpu-h200-deepep
 
   unit-test-backend-4-gpu-b200:
-    needs: [check-changes, call-gate, unit-test-backend-2-gpu]
+    needs: [check-changes, call-gate, unit-test-backend-2-gpu, stage-b-test-4-gpu-b200]
     if: |
       always() &&
       (
@@ -1229,7 +1271,7 @@ jobs:
           IS_BLACKWELL=1 python3 run_suite.py --suite per-commit-4-gpu-b200 --auto-partition-id ${{ matrix.part }}  --auto-partition-size 3 --timeout-per-file 1800
 
   unit-test-backend-4-gpu-gb200:
-    needs: [check-changes, call-gate, unit-test-backend-2-gpu, sgl-kernel-build-wheels-arm]
+    needs: [check-changes, call-gate, unit-test-backend-2-gpu, stage-b-test-4-gpu-b200, sgl-kernel-build-wheels-arm]
     if: |
       always() &&
       (
@@ -1287,6 +1329,7 @@ jobs:
         quantization-test,
         unit-test-backend-1-gpu,
         unit-test-backend-2-gpu,
+        stage-b-test-4-gpu-b200,
         unit-test-backend-4-gpu,
         unit-test-backend-8-gpu-h20,
         unit-test-backend-8-gpu-h200,
diff --git a/scripts/ci/slash_command_handler.py b/scripts/ci/slash_command_handler.py
index bb09d52a5..fa3bb8557 100644
--- a/scripts/ci/slash_command_handler.py
+++ b/scripts/ci/slash_command_handler.py
@@ -149,6 +149,7 @@ def handle_rerun_stage(
         "quantization-test",
         "unit-test-backend-1-gpu",
         "unit-test-backend-2-gpu",
+        "stage-b-test-4-gpu-b200",
         "unit-test-backend-4-gpu",
         "unit-test-backend-8-gpu-h200",
         "unit-test-backend-8-gpu-h20",
diff --git a/scripts/ci_monitor/ci_analyzer.py b/scripts/ci_monitor/ci_analyzer.py
index 05bbab28c..409960f16 100755
--- a/scripts/ci_monitor/ci_analyzer.py
+++ b/scripts/ci_monitor/ci_analyzer.py
@@ -105,6 +105,7 @@ class SGLangCIAnalyzer:
                 "stage-a-test-1",
                 "unit-test-backend-1-gpu",
                 "unit-test-backend-2-gpu",
+                "stage-b-test-4-gpu-b200",
                 "unit-test-backend-4-gpu",
                 "unit-test-backend-8-gpu",
             ],
@@ -213,6 +214,7 @@ class SGLangCIAnalyzer:
                     "stage-a-test-1",
                     "unit-test-backend-1-gpu",
                     "unit-test-backend-2-gpu",
+                    "stage-b-test-4-gpu-b200",
                     "unit-test-backend-4-gpu",
                     "unit-test-backend-8-gpu-h200",
                     "unit-test-backend-8-gpu-h20",
diff --git a/scripts/ci_monitor/ci_analyzer_balance.py b/scripts/ci_monitor/ci_analyzer_balance.py
index e0779990c..9217b3c81 100755
--- a/scripts/ci_monitor/ci_analyzer_balance.py
+++ b/scripts/ci_monitor/ci_analyzer_balance.py
@@ -177,6 +177,7 @@ class SGLangTestBalanceAnalyzer:
             "stage-a-test-1",
             "unit-test-backend-1-gpu",
             "unit-test-backend-2-gpu",
+            "stage-b-test-4-gpu-b200",
             "unit-test-backend-4-gpu",
             "unit-test-backend-8-gpu-h200",
             "unit-test-backend-8-gpu-h20",
diff --git a/test/srt/run_suite.py b/test/srt/run_suite.py
index 52389e8c2..ca84c0612 100644
--- a/test/srt/run_suite.py
+++ b/test/srt/run_suite.py
@@ -167,8 +167,10 @@ suites = {
         TestFile("test_disaggregation_pp.py", 140),
         TestFile("test_disaggregation_dp_attention.py", 155),
     ],
+    "per-commit-4-gpu-b200-stage-b": [
+        TestFile("test_deepseek_v3_fp4_4gpu.py", 1800),  # Stage B test
+    ],
     "per-commit-4-gpu-b200": [
-        TestFile("test_deepseek_v3_fp4_4gpu.py", 1800),
         TestFile("test_flash_attention_4.py", 90),
         TestFile("test_fp8_blockwise_gemm.py", 280),
         TestFile("test_gpt_oss_4gpu.py", 700),