diff --git a/python/sglang/srt/entrypoints/grpc_bridge.py b/python/sglang/srt/entrypoints/grpc_bridge.py
index dbab46081..dc62075bb 100644
--- a/python/sglang/srt/entrypoints/grpc_bridge.py
+++ b/python/sglang/srt/entrypoints/grpc_bridge.py
@@ -299,7 +299,9 @@ class RuntimeHandle:
             gen = self.tokenizer_manager.generate_request(obj, request=request)
             if stream:
                 completed_choices = set()
-                expected_choices = obj.batch_size * obj.parallel_sample_num
+                # generate_request does not normalize obj until iteration begins.
+                sampling_params = obj.sampling_params or {}
+                expected_choices = max(1, int(sampling_params.get("n", 1)))
                 async for chunk in gen:
                     choice_finished = (
                         chunk.get("meta_info", {}).get("finish_reason") is not None
diff --git a/test/registered/unit/entrypoints/test_grpc_bridge.py b/test/registered/unit/entrypoints/test_grpc_bridge.py
index dccc7b24b..6aa741bb8 100644
--- a/test/registered/unit/entrypoints/test_grpc_bridge.py
+++ b/test/registered/unit/entrypoints/test_grpc_bridge.py
@@ -93,7 +93,7 @@ class TestNativeGrpcParallelResponses(CustomTestCase):
             },
         ]
         handle = _make_runtime_handle(responses)
-        obj = SimpleNamespace(rid="logical", batch_size=1, parallel_sample_num=2)
+        obj = SimpleNamespace(rid="logical", sampling_params={"n": 2})
 
         asyncio.run(
             handle._run_generate(