fix(azure): strip the gpt5_series/ selector from Responses deployment names

A deployment configured as azure/gpt5_series/<deployment> sent gpt5_series/<deployment> as the model of Azure Responses requests, which Azure rejects because no such deployment exists. Chat completions already strip the selector, so Responses now does the same.

litellm/llms/azure/responses/transformation.py

@@ -110,7 +110,8 @@ class AzureOpenAIResponsesAPIConfig(OpenAIResponsesAPIConfig):
         return BaseAzureLLM.base_validate_azure_environment(headers=headers, litellm_params=litellm_params)
 
     def get_stripped_model_name(self, model: str) -> str:
-        return model.replace("responses/", "").replace("o_series/", "").replace("azure_ai/", "")
+        stripped: Final = model.replace("responses/", "").replace("o_series/", "").replace("azure_ai/", "")
+        return stripped.removeprefix(AzureOpenAIGPT5Config.GPT5_SERIES_ROUTE)
 
     def _handle_reasoning_item(self, item: dict[str, Any]) -> Mapping[str, object]:
         """

tests/unit/llms/azure/response/test_azure_transformation.py

@@ -1297,6 +1297,23 @@ def test_azure_responses_sends_the_deployment_name_when_azure_ai_prefix_survives
     assert request["model"] == "gpt-5.4-nano"
 
 
+def test_gpt5_series_selector_is_not_sent_as_part_of_the_deployment_name(respx_mock: respx.MockRouter) -> None:
+    route: Final = respx_mock.post(host="example-resource.openai.azure.com").mock(
+        return_value=httpx.Response(200, json=_COMPLETED_RESPONSE)
+    )
+
+    response: Final = litellm.responses(
+        model="azure/gpt5_series/my-gpt-5-deployment",
+        input="hi",
+        api_base=_AZURE_RESOURCE,
+        api_key="azure-test-key",
+    )
+
+    assert response.output[0].content[0].text == "Done."
+    assert route.call_count == 1
+    assert json.loads(route.calls[0].request.content)["model"] == "my-gpt-5-deployment"
+
+
 @pytest.mark.asyncio
 async def test_azure_responses_api_status_error():
     """