fix(azure): send the deployment name as the body model on v1 image routes

litellm/llms/azure/azure.py

@@ -846,6 +846,7 @@ class AzureChatCompletion(BaseAzureLLM, BaseLLM):
         api_key: str,
         data: dict,
         headers: dict,
+        deployment_name: str | None = None,
     ) -> httpx.Response:
         """
         Implemented for azure dall-e-2 image gen calls
@@ -957,7 +958,7 @@ class AzureChatCompletion(BaseAzureLLM, BaseLLM):
                 content=json.dumps(result).encode("utf-8"),
                 request=httpx.Request(method="POST", url="https://api.openai.com/v1"),
             )
-        request_json: Final = azure_deployment_image_generation_json_body(api_base, data)
+        request_json: Final = azure_deployment_image_generation_json_body(api_base, data, deployment_name)
         return await async_handler.post(
             url=api_base,
             json=request_json,
@@ -973,6 +974,7 @@ class AzureChatCompletion(BaseAzureLLM, BaseLLM):
         api_key: str,
         data: dict,
         headers: dict,
+        deployment_name: str | None = None,
     ) -> httpx.Response:
         """
         Implemented for azure dall-e-2 image gen calls
@@ -1073,7 +1075,7 @@ class AzureChatCompletion(BaseAzureLLM, BaseLLM):
                 content=json.dumps(result).encode("utf-8"),
                 request=httpx.Request(method="POST", url="https://api.openai.com/v1"),
             )
-        request_json: Final = azure_deployment_image_generation_json_body(api_base, data)
+        request_json: Final = azure_deployment_image_generation_json_body(api_base, data, deployment_name)
         return sync_handler.post(
             url=api_base,
             json=request_json,
@@ -1176,6 +1178,7 @@ class AzureChatCompletion(BaseAzureLLM, BaseLLM):
                 api_key=api_key,
                 data=data,
                 headers=headers,
+                deployment_name=model,
             )
 
             provider_config: Final = get_azure_image_generation_config(data.get("model", "dall-e-2"))
@@ -1311,6 +1314,7 @@ class AzureChatCompletion(BaseAzureLLM, BaseLLM):
                 api_key=api_key or "",
                 data=data,
                 headers=headers,
+                deployment_name=model,
             )
             provider_config: Final = get_azure_image_generation_config(data.get("model", "dall-e-2"))
             if isinstance(provider_config, AzureFoundryMAIImageGenerationConfig):

litellm/llms/azure/image_generation/http_utils.py

@@ -1,7 +1,9 @@
 """HTTP helpers for Azure OpenAI image generation (REST, not SDK)."""
 
+from typing import Final
 
-def azure_deployment_image_generation_json_body(api_base: str, data: dict) -> dict:
+
+def azure_deployment_image_generation_json_body(api_base: str, data: dict, deployment_name: str | None = None) -> dict:
     """
     Build the JSON body for Azure OpenAI image generation POSTs.
 
@@ -9,9 +11,20 @@ def azure_deployment_image_generation_json_body(api_base: str, data: dict) -> di
     deployment in the URL only; sending ``model`` in the body (especially the deployment
     name) breaks some models (e.g. gpt-image-2). See LiteLLM #26316.
 
+    For the v1 surface (``.../openai/v1/images/...``), Azure routes by the deployment
+    name in the body ``model`` field, so the deployment name must replace any base
+    model name there or Azure answers 404 DeploymentNotFound.
+
     Provider-style URLs (e.g. ``/providers/...`` for FLUX on Azure AI) keep all keys
     so non–OpenAI-deployment payloads still work.
     """
-    if "images/generations" in api_base and "/openai/deployments/" in api_base:
-        return {k: v for k, v in data.items() if k != "model"}
-    return data
+    drop_model: Final = "images/generations" in api_base and "/openai/deployments/" in api_base
+    v1_route: Final = "/openai/v1/images/" in api_base and bool(deployment_name)
+    if not drop_model and not v1_route:
+        return data
+    entries: Final = (
+        tuple((k, v) for k, v in data.items() if k != "model")
+        if drop_model
+        else (*data.items(), ("model", deployment_name))
+    )
+    return {k: v for k, v in entries}

tests/test_litellm/llms/azure/image_generation/test_azure_image_generation_init.py

@@ -3,9 +3,12 @@ import traceback
 from typing import Callable, Optional
 from unittest.mock import AsyncMock, MagicMock, Mock, patch
 
+import httpx
 import pytest
+import respx
 
 import litellm
+from litellm.caching.llm_caching_handler import LLMClientCache
 from litellm.llms.azure.azure import AzureChatCompletion
 from litellm.llms.azure.image_generation.http_utils import (
     azure_deployment_image_generation_json_body,
@@ -489,3 +492,98 @@ def test_azure_image_generation_v1_api_version_uses_base_url_client_param():
         base_model=None,
     )
     assert url == "https://my-resource.openai.azure.com/openai/v1/images/generations?api-version=preview"
+
+
+def test_azure_v1_image_generation_json_body_sends_deployment_name():
+    """The v1 route ignores the URL and routes by body ``model``, which must be the deployment name."""
+    url = "https://my-resource.openai.azure.com/openai/v1/images/generations?api-version=preview"
+    data = {"model": "gpt-image-2", "prompt": "x", "n": 1}
+    out = azure_deployment_image_generation_json_body(url, data, deployment_name="img-dep")
+    assert out["model"] == "img-dep"
+    assert out["prompt"] == "x"
+    assert data["model"] == "gpt-image-2"
+    assert azure_deployment_image_generation_json_body(url, data) == data
+
+
+@pytest.mark.asyncio
+async def test_azure_aimage_generation_v1_route_sends_deployment_name_in_body(
+    respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch
+):
+    monkeypatch.setattr(litellm, "disable_aiohttp_transport", True)
+    monkeypatch.setattr(litellm, "in_memory_llm_clients_cache", LLMClientCache())
+    azure_chat_completion = AzureChatCompletion()
+    model = "img-dep"
+    base_model = "gpt-image-2"
+    data = {"model": base_model, "prompt": "A beautiful image of a cat", "n": 1}
+    azure_client_params = {
+        "azure_endpoint": "https://my-resource.openai.azure.com",
+        "api_version": "preview",
+    }
+
+    route = respx_mock.post("https://my-resource.openai.azure.com/openai/v1/images/generations").mock(
+        return_value=httpx.Response(200, json={"created": 1234567890, "data": [{"b64_json": "aaaa"}]})
+    )
+
+    logging_obj = MagicMock()
+    logging_obj.pre_call = MagicMock()
+    logging_obj.post_call = MagicMock()
+
+    await azure_chat_completion.aimage_generation(
+        data=data,
+        model_response=None,
+        azure_client_params=azure_client_params,
+        api_key="test-api-key",
+        input=[],
+        logging_obj=logging_obj,
+        headers={},
+        model=model,
+        timeout=60.0,
+    )
+
+    request = route.calls.last.request
+    assert str(request.url) == ("https://my-resource.openai.azure.com/openai/v1/images/generations?api-version=preview")
+    sent_body = json.loads(request.content)
+    assert sent_body["model"] == model
+    assert sent_body["prompt"] == data["prompt"]
+
+
+def test_azure_image_generation_v1_route_base_model_vs_deployment_name(respx_mock: respx.MockRouter):
+    """On the v1 surface the body ``model`` must be the deployment name, never base_model."""
+    azure_chat_completion = AzureChatCompletion()
+    prompt = "A beautiful image of a cat"
+    model = "img-dep"
+    base_model = "gpt-image-2"
+    api_base = "https://my-resource.openai.azure.com"
+    api_version = "v1"
+    litellm_params = {
+        "base_model": base_model,
+        "api_base": api_base,
+        "api_version": api_version,
+    }
+
+    route = respx_mock.post(f"{api_base}/openai/v1/images/generations").mock(
+        return_value=httpx.Response(200, json={"created": 1234567890, "data": [{"b64_json": "aaaa"}]})
+    )
+
+    logging_obj = MagicMock()
+    logging_obj.pre_call = MagicMock()
+    logging_obj.post_call = MagicMock()
+
+    azure_chat_completion.image_generation(
+        prompt=prompt,
+        timeout=60.0,
+        optional_params={"n": 1, "size": "1024x1024"},
+        logging_obj=logging_obj,
+        headers={},
+        model=model,
+        api_key="test-api-key",
+        api_base=api_base,
+        api_version=api_version,
+        litellm_params=litellm_params,
+    )
+
+    request = route.calls.last.request
+    assert str(request.url) == f"{api_base}/openai/v1/images/generations?api-version={api_version}"
+    sent_body = json.loads(request.content)
+    assert sent_body["model"] == model
+    assert sent_body["prompt"] == prompt