fix: reject failed and incomplete Responses statuses in non-streaming path

src/agents/extensions/models/any_llm_model.py

@@ -360,6 +360,13 @@ async def _get_response_via_responses(
                 prompt=prompt,
             )
 
+            status = getattr(response, "status", None)
+            if status in {"failed", "incomplete"}:
+                raise response_terminal_failure_error(
+                    f"response.{status}",
+                    response if isinstance(response, Response) else None,
+                )
+
             if _debug.DONT_LOG_MODEL_DATA:
                 logger.debug("LLM responded")
             else:

src/agents/models/openai_responses.py

@@ -472,6 +472,13 @@ async def get_response(
                     prompt=prompt,
                 )
 
+                status = getattr(response, "status", None)
+                if status in {"failed", "incomplete"}:
+                    raise response_terminal_failure_error(
+                        f"response.{status}",
+                        response if isinstance(response, Response) else None,
+                    )
+
                 if _debug.DONT_LOG_MODEL_DATA:
                     logger.debug("LLM responded")
                 else:

tests/models/test_any_llm_model.py

@@ -623,6 +623,35 @@ async def response_stream() -> AsyncIterator[ResponseErrorEvent]:
     assert events[0].code == "invalid_request_error"
 
 
+@pytest.mark.allow_call_model_methods
+@pytest.mark.asyncio
+@pytest.mark.parametrize("status", ["failed", "incomplete"])
+async def test_any_llm_responses_path_rejects_failed_or_incomplete_status(
+    monkeypatch,
+    status: str,
+) -> None:
+    response = _response("partial", response_id="resp-terminal")
+    response.status = status  # type: ignore[assignment]
+    provider = FakeAnyLLMProvider(supports_responses=True, responses_response=response)
+    module, _create_calls = _import_any_llm_module(monkeypatch, provider)
+    AnyLLMModel = module.AnyLLMModel
+
+    model = AnyLLMModel(model="openai/gpt-5.4-mini")
+    with pytest.raises(ModelBehaviorError, match=f"response.{status}"):
+        await model.get_response(
+            system_instructions=None,
+            input="hi",
+            model_settings=ModelSettings(),
+            tools=[],
+            output_schema=None,
+            handoffs=[],
+            tracing=ModelTracing.DISABLED,
+            previous_response_id=None,
+            conversation_id=None,
+            prompt=None,
+        )
+
+
 @pytest.mark.allow_call_model_methods
 @pytest.mark.asyncio
 async def test_any_llm_responses_path_passes_transport_kwargs_via_private_provider_api(

tests/models/test_openai_responses.py

@@ -312,6 +312,34 @@ def __init__(self):
     }
 
 
+@pytest.mark.allow_call_model_methods
+@pytest.mark.asyncio
+@pytest.mark.parametrize("status", ["failed", "incomplete"])
+async def test_get_response_rejects_failed_or_incomplete_response_status(status: str) -> None:
+    class DummyResponses:
+        async def create(self, **kwargs: Any) -> Any:
+            response = get_response_obj([], response_id="resp-terminal")
+            response.status = status  # type: ignore[assignment]
+            return response
+
+    class DummyResponsesClient:
+        def __init__(self) -> None:
+            self.responses = DummyResponses()
+
+    model = OpenAIResponsesModel(model="gpt-4", openai_client=DummyResponsesClient())  # type: ignore[arg-type]
+
+    with pytest.raises(ModelBehaviorError, match=f"response.{status}"):
+        await model.get_response(
+            system_instructions=None,
+            input="hi",
+            model_settings=ModelSettings(),
+            tools=[],
+            output_schema=None,
+            handoffs=[],
+            tracing=ModelTracing.DISABLED,
+        )
+
+
 def test_get_client_disables_provider_managed_retries_on_runner_retry() -> None:
     class DummyResponsesClient:
         def __init__(self) -> None: