From ada0a1ad3a06d3ae970a0e5e61d224867fb2fcf5 Mon Sep 17 00:00:00 2001
From: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
Date: Wed, 16 Sep 2026 16:49:29 -0700
Subject: [PATCH] fix(azure_ai): strip the azure_ai/ prefix when a Responses
 call is remapped to azure

A catalog OpenAI name on an .openai.azure.com host (or with AZURE_AI_API_BASE set to one) is remapped from azure_ai to azure before the Responses request is built, and the azure_ai/ prefix stayed in the wire model, so Azure answered DeploymentNotFound. The Azure Responses config now strips azure_ai/ next to responses/ and o_series/.
---
 litellm/llms/azure/responses/transformation.py |  7 +------
 .../response/test_azure_transformation.py      | 11 +++++++++++
 .../test_azure_ai_responses_transformation.py  | 18 ++++++++++++++++++
 3 files changed, 30 insertions(+), 6 deletions(-)

diff --git a/litellm/llms/azure/responses/transformation.py b/litellm/llms/azure/responses/transformation.py
index 7fe12138ebc..2a82b42df7b 100644
--- a/litellm/llms/azure/responses/transformation.py
+++ b/litellm/llms/azure/responses/transformation.py
@@ -49,12 +49,7 @@ class AzureOpenAIResponsesAPIConfig(OpenAIResponsesAPIConfig):
         return BaseAzureLLM._base_validate_azure_environment(headers=headers, litellm_params=litellm_params)
 
     def get_stripped_model_name(self, model: str) -> str:
-        # if "responses/" is in the model name, remove it
-        if "responses/" in model:
-            model = model.replace("responses/", "")
-        if "o_series" in model:
-            model = model.replace("o_series/", "")
-        return model
+        return model.replace("responses/", "").replace("o_series/", "").replace("azure_ai/", "")
 
     def _handle_reasoning_item(self, item: dict[str, Any]) -> dict[str, Any]:
         """
diff --git a/tests/test_litellm/llms/azure/response/test_azure_transformation.py b/tests/test_litellm/llms/azure/response/test_azure_transformation.py
index 726c9f65681..532c278e891 100644
--- a/tests/test_litellm/llms/azure/response/test_azure_transformation.py
+++ b/tests/test_litellm/llms/azure/response/test_azure_transformation.py
@@ -677,3 +677,14 @@ def test_azure_responses_gpt6_astra_rejects_temperature_while_reasoning(local_mo
             model="gpt-6-astra",
             drop_params=False,
         )
+
+
+def test_azure_responses_sends_the_deployment_name_when_azure_ai_prefix_survives_provider_remap():
+    request = AzureOpenAIResponsesAPIConfig().transform_responses_api_request(
+        model="azure_ai/gpt-5.4-nano",
+        input="hi",
+        response_api_optional_request_params={},
+        litellm_params=GenericLiteLLMParams(),
+        headers={},
+    )
+    assert request["model"] == "gpt-5.4-nano"
diff --git a/tests/test_litellm/llms/azure_ai/responses/test_azure_ai_responses_transformation.py b/tests/test_litellm/llms/azure_ai/responses/test_azure_ai_responses_transformation.py
index 925608a3c9b..bae956eb061 100644
--- a/tests/test_litellm/llms/azure_ai/responses/test_azure_ai_responses_transformation.py
+++ b/tests/test_litellm/llms/azure_ai/responses/test_azure_ai_responses_transformation.py
@@ -253,6 +253,24 @@ async def test_aresponses_sends_reasoning_and_tools_to_native_endpoint(model, ap
     _assert_native_responses_request(route, expected_url, expected_model)
 
 
+@pytest.mark.asyncio
+@respx.mock
+async def test_aresponses_catalog_name_remapped_to_azure_sends_bare_deployment_name(monkeypatch):
+    monkeypatch.setenv("AZURE_AI_API_BASE", "https://res.openai.azure.com")
+    route = respx.post(url__regex=r".*/openai/v1/responses(\?.*)?$").mock(
+        return_value=httpx.Response(200, json=_responses_payload("gpt-5.4-nano"))
+    )
+
+    await litellm.aresponses(
+        model="azure_ai/gpt-5.4-nano",
+        input="What is the weather in SF?",
+        api_base="https://res.openai.azure.com",
+        api_key="fake-key",
+    )
+
+    assert json.loads(route.calls.last.request.content)["model"] == "gpt-5.4-nano"
+
+
 @pytest.mark.asyncio
 @respx.mock
 @pytest.mark.parametrize("model,api_base,expected_url,expected_model", NATIVE_RESPONSES_CASES)