GPT 5 not working with the azure pipline

[Breidenbach-Murach]

I tried implementing the new gpt-5 model after we got access from Azure today.
With the default Azure AI Foundry pipeline we're getting the error
Error: Unsupported parameter: 'messages'. In the Responses API, this parameter has moved to 'input'. Try again with the new parameter. See the API documentation for more information: https://platform.openai.com/docs/api-reference/responses/create .

So I tried to change messages to input and added a very crude: body["input"] = body["messages"] here

That takes care of the error, but I'm not getting any response in the chat window now, even though it says it's streaming a response for a few seconds.
Since I don't really know how to proceed debugging this further, I thought I'd ask for some help. 😅

[owndev]

Hi @Breidenbach-Murach

I don't have access to GPT-5. I could test it only when it's open to everyone. But that could take a long time 😅.
I've tested the other GPT-5 models (gpt-5-mini, gpt-5-nano, gpt-5-chat) and there were no issues. You might want to use one of those models.
Or go directly via the OpenAI API instead of Azure.

[Breidenbach-Murach]

Hey,

I just tried deploying gpt-5-mini and I'm getting the same error.

open-webui is on the latest version v0.6.22 and I'm using a fresh pipeline with version 2.3.1.
Disabled all other functions, but it's still happening.

The open-webui log has the following error:

2025-08-12 07:25:36.853 | ERROR    | function_gpt_5_mini:pipe:570 - Error in Azure AI request: 400, message='Bad Request', url='https://{endpoint}.openai.azure.com/openai/responses?api-version=2025-04-01-preview'
Traceback (most recent call last):

  File "<frozen runpy>", line 198, in _run_module_as_main
  File "<frozen runpy>", line 88, in _run_code
  File "/usr/local/lib/python3.11/site-packages/uvicorn/__main__.py", line 4, in <module>
    uvicorn.main()
    │       └ <Command main>
    └ <module 'uvicorn' from '/usr/local/lib/python3.11/site-packages/uvicorn/__init__.py'>
  File "/usr/local/lib/python3.11/site-packages/click/core.py", line 1442, in __call__
    return self.main(*args, **kwargs)
           │    │     │       └ {}
           │    │     └ ()
           │    └ <function Command.main at 0x7c4170ca6ac0>
           └ <Command main>
  File "/usr/local/lib/python3.11/site-packages/click/core.py", line 1363, in main
    rv = self.invoke(ctx)
         │    │      └ <click.core.Context object at 0x7c4171a2ca50>
         │    └ <function Command.invoke at 0x7c4170ca67a0>
         └ <Command main>
  File "/usr/local/lib/python3.11/site-packages/click/core.py", line 1226, in invoke
    return ctx.invoke(self.callback, **ctx.params)
           │   │      │    │           │   └ {'host': '0.0.0.0', 'port': 8080, 'forwarded_allow_ips': '*', 'workers': 1, 'app': 'open_webui.main:app', 'uds': None, 'fd': ...
           │   │      │    │           └ <click.core.Context object at 0x7c4171a2ca50>
           │   │      │    └ <function main at 0x7c4170962de0>
           │   │      └ <Command main>
           │   └ <function Context.invoke at 0x7c4170ca59e0>
           └ <click.core.Context object at 0x7c4171a2ca50>
  File "/usr/local/lib/python3.11/site-packages/click/core.py", line 794, in invoke
    return callback(*args, **kwargs)
           │         │       └ {'host': '0.0.0.0', 'port': 8080, 'forwarded_allow_ips': '*', 'workers': 1, 'app': 'open_webui.main:app', 'uds': None, 'fd': ...
           │         └ ()
           └ <function main at 0x7c4170962de0>
  File "/usr/local/lib/python3.11/site-packages/uvicorn/main.py", line 413, in main
    run(
    └ <function run at 0x7c4170d87b00>
  File "/usr/local/lib/python3.11/site-packages/uvicorn/main.py", line 580, in run
    server.run()
    │      └ <function Server.run at 0x7c4170c08c20>
    └ <uvicorn.server.Server object at 0x7c4170b8bb50>
  File "/usr/local/lib/python3.11/site-packages/uvicorn/server.py", line 67, in run
    return asyncio.run(self.serve(sockets=sockets))
           │       │   │    │             └ None
           │       │   │    └ <function Server.serve at 0x7c4170c08cc0>
           │       │   └ <uvicorn.server.Server object at 0x7c4170b8bb50>
           │       └ <function run at 0x7c41710d9440>
           └ <module 'asyncio' from '/usr/local/lib/python3.11/asyncio/__init__.py'>
  File "/usr/local/lib/python3.11/asyncio/runners.py", line 190, in run
    return runner.run(main)
           │      │   └ <coroutine object Server.serve at 0x7c4170930a90>
           │      └ <function Runner.run at 0x7c4170f4cfe0>
           └ <asyncio.runners.Runner object at 0x7c417096b8d0>
  File "/usr/local/lib/python3.11/asyncio/runners.py", line 118, in run
    return self._loop.run_until_complete(task)
           │    │     │                  └ <Task pending name='Task-1' coro=<Server.serve() running at /usr/local/lib/python3.11/site-packages/uvicorn/server.py:71> wai...
           │    │     └ <cyfunction Loop.run_until_complete at 0x7c41709b05f0>
           │    └ <uvloop.Loop running=True closed=False debug=False>
           └ <asyncio.runners.Runner object at 0x7c417096b8d0>

  File "/app/backend/open_webui/utils/middleware.py", line 2155, in response_handler
    await stream_body_handler(response, form_data)
          │                   │         └ {'stream': True, 'model': 'gpt_5_mini.gpt-5-mini', 'messages': [{'role': 'user', 'content': 'This is a test.'}]}
          │                   └ <starlette.responses.StreamingResponse object at 0x7c412bb36d90>
          └ <function process_chat_response.<locals>.response_handler.<locals>.stream_body_handler at 0x7c4121f34400>

  File "/app/backend/open_webui/utils/middleware.py", line 1850, in stream_body_handler
    async for line in response.body_iterator:
                      │        └ <async_generator object generate_function_chat_completion.<locals>.stream_content at 0x7c412b7ab760>
                      └ <starlette.responses.StreamingResponse object at 0x7c412bb36d90>

  File "/app/backend/open_webui/functions.py", line 270, in stream_content
    res = await execute_pipe(pipe, params)
                │            │     └ {'body': {'stream': True, 'model': 'gpt_5_mini.gpt-5-mini', 'messages': [{'role': 'user', 'content': 'This is a test.'}]}, '_...
                │            └ <bound method Pipe.pipe of <function_gpt_5_mini.Pipe object at 0x7c4121ee5790>>
                └ <function generate_function_chat_completion.<locals>.execute_pipe at 0x7c412bbb14e0>

  File "/app/backend/open_webui/functions.py", line 141, in execute_pipe
    return await pipe(**params)
                 │      └ {'body': {'stream': True, 'model': 'gpt_5_mini.gpt-5-mini', 'messages': [{'role': 'user', 'content': 'This is a test.'}]}, '_...
                 └ <bound method Pipe.pipe of <function_gpt_5_mini.Pipe object at 0x7c4121ee5790>>

> File "<string>", line 556, in pipe
  File "/usr/local/lib/python3.11/site-packages/aiohttp/client_reqrep.py", line 629, in raise_for_status
    raise ClientResponseError(
          └ <class 'aiohttp.client_exceptions.ClientResponseError'>

aiohttp.client_exceptions.ClientResponseError: 400, message='Bad Request', url='https://{endpoint}.openai.azure.com/openai/responses?api-version=2025-04-01-preview'

The only difference to the currently working gpt-4.1, gpt-4o and gpt-4o-mini I can tell are the API versions in the endpoint.
The old ones use 2025-01-01-preview, whereas the new gpt-5 and gpt-5-mini use 2025-04-01-preview.

Anything else I can try, or any more info I can provide?
If required, I could provide you with an API key for testing.

P.S.: I don't think this is significant, but we're using the Sweden Central Azure region for our AI Foundry Assets.

[owndev]

I see the difference between our configurations. I use: https://{endpoint}.services.ai.azure.com/models/chat/completions?api-version=2024-05-01-preview and you use https://{endpoint}.openai.azure.com/openai/responses?api-version=2025-04-01-preview.

My endpoint can use all Azure models, whereas yours only applies to OpenAI models. Currently, the endpoint /openai/responses is not supported because it works differently.

Try this endpoint to see if it works: https://{endpoint}.services.ai.azure.com/models/chat/completions?api-version=2024-05-01-preview

[Breidenbach-Murach]

The other endpoint is working for both models.
Thanks for the insight and quick help!