Azure OpenAI Assistants suddenly not working properly (both Playground & API)

Question

Azure OpenAI Assistants suddenly not working properly (both Playground & API)

Tae-Gil Noh 30

It seems that something is wrong with recent model roll-outs, affecting Azure OpenAI Assistants.

This is easy to reproduce:

Visit Azure AI Foundry -> Azure OpenAI Service -> Playground -> Assistant Playground
Create a new Assistant on regions where Assistant preview is available (e.g. sweden, or us east region where Assistant is supported.)
Error case 1) Try to change any parameter (e.g. temperature) --> update will fail with strange error message such as "temperature is not supported"
Error case 2) try any simple message with the assistant --> this will fail with strange message about either "reasoning_effort", or "stream"
Note that
1. All existing Assistants are affected too -> same errors occur
2. The results are same also for REST API calls and SDK calls (tested with JAVA SDK beta.5 version)

Ajouz, Ali 5 Reputation points

2025-02-27T13:57:25.5833333+00:00

I can confirm this problem. It impacts existing deployments
Tae-Gil Noh 30 Reputation points

2025-02-28T13:54:32.99+00:00
I think I now have more information about why some users this occurs, and why some users are okay with the Assistant.

If the Assistant is created within OpenAI resource ==> does NOT work (both new and existing)

If the Assistant is created within AIServices as "Hub" ==> does work (both new and existing)

As an example, the following picture:

AIServices "ai-useast1testhub59971" --> Assistants works okay there, no problem.

OpenAI "omq-gpt-models-us-ea..." --> Assistant does NOT work at all, as of now.

Please note that, "AIServices" resource is a "AI Hub" which has been created with "Create Project" process of AI Foundary. "OpenAI" resource is a service resource created from Azure site (just like creating any service). As I see it now, Assistants are currently working only on "AIServices" type of resource, but not with "OpenAI" resource.

To the Azure or Azure vendor personal who reads this -- this is still a terrible bug that breaks many running and pre-setup OpenAI Assistants running on Azure. Please fix this ASAP. I hope the above investigation was helpful to report / communicate the bug to AOAI dev team.
Tae-Gil Noh 30 Reputation points

2025-02-28T13:56:35.77+00:00

This is still a bug, IMHO. Please fix this ASAP!!
Marc Rousset 0 Reputation points

2025-03-03T13:53:59.13+00:00

I've been stuck on this and unable to resolve - either using OpenAI or AI Services, new and old subs. Which region are you in?
Tae-Gil Noh 30 Reputation points

2025-03-04T10:48:39.37+00:00
In my case, I was able to let it work on US east2 and Sweden central.

Maybe useful info

Both assistant and GPT created within the AIService Hub. (Linking existing resource to OpenAI didn't work).

All API calls directed to hub URL (that has microsoft cognition URLs)

The only way that I could make a hub that works with OpenAI Assistant was create a new project in AI Foundry with create new (new resource, new source group, new everything) and use that newly created hub. When I tried to connect it to previously existing resource, I've got the same errors....

However, I am not sure if this is a good workaround at all. e.g. maybe OpenAI service is running latest code, and some AI services code is not updated yet and working just for the reason -- who knows. It seems that the "preview" state of Azure OpenAI is much more volatile than "preview" state of OpenAI itself. sigh.
santoshkc 15,355 Reputation points Microsoft External Staff Moderator

2025-03-04T13:08:18.0433333+00:00

Hi @Tae-Gil Noh,

Apologies for the delay in response and thank you for reaching out and sharing these detailed observations. It looks like the issue is related to recent model updates affecting Assistants created under OpenAI resources, while those under AI Services (AI Hub) continue to function correctly. The errors around "temperature" and "reasoning_effort" suggest that certain parameters may no longer be supported for specific models.

We appreciate your insights into the difference between OpenAI resources and AI Service Hubs. If AI Hubs are working as expected, it could indicate a temporary inconsistency in how updates are being applied across different Azure environments. We recommend continuing with AI Hub-based Assistants for now while we investigate this further. Additionally, we are working on this internally to understand whether this is an expected change or a temporary issue. And please let us know if you notice any further changes or if you require additional assistance.

Thank you.
santoshkc 15,355 Reputation points Microsoft External Staff Moderator

2025-03-05T16:13:25.36+00:00

Hi @Tae-Gil Noh,

We haven’t heard from you on the last response and was just checking back to see if the given information was helpful.

Thank you.
Tae-Gil Noh 30 Reputation points

2025-03-05T18:33:54.3833333+00:00
@santoshkc your comment was not helping at all.

Azure OpenAI Assistant is still not working

Assistants, in Azure OpenAI Service, are still broken -- just tested at March 5th CET 19:20.

You didn't give us any additional information about this -- why? how? what are work arounds? when will this be fixed?

If you are an official support personal --- please be decent and provide us information like the followings

what is the current state of the problem?

is Azure OpenAI development team working for the issue already? e.g. a work ticket already out there?

when will it be fixed?

is there a solid and reliable work around? I've tried some, but as you see from the comment of others, that doesn't seem to work well for some others.

Your comment did not add anything about the issue, and basically just repeated what I have posted.

And Now you are asking : "... just checking back to see if the given information was helpful."

Sadly No, not at all:

a) We are still waiting for the issue to be fixed

b) If that takes time -- we would like to know when it will be fixed

c) If even this timeframe is not sure, then we would like to know minimally -- what is a reliable workaround for us to use Assistants APIs in a reliable way?

Thank you. We would really love to have information and/or solutions.
Tae-Gil Noh 30 Reputation points

2025-03-06T08:03:39.0933333+00:00
March 6th, 09:00 CET -- it still fails.

Posting the minimal reproducable sequence and error codes, just in case this helps Azure support to trace the issue.

Create a new Assistant on sweden-central region, on existing OpenAI service, via Assistants playground, with an existing gpt4-o (version 2024-08-06)

it doesn't matter what GPT model, version, or newly created or not -- all fails

Give "hello" to the assistant -- normally there should be a response, but here an error:

400: Unsupported value: 'stream' does not support 'true' with this model. Supported values are: 'false'. | Apim-request-id: f0e8fe72-4a73-41d9-a583-194fa264008f

Nothing works on the Assistant, for example, tried to update configuration of the Assistant, such as temperature value

Failed to update assistant ``unsupported_model: Unsupported parameter: 'temperature' is not supported with this model. Trace ID : b616aa37-2615-467f-ae8d-735a3970742cClient request ID : 3be1289d-98ed-4050-9c6e-dfcd82eff123APIM request ID : 4e89edca-ea14-4cbb-abf5-71bfc84e772b

Note that, the above error messages are not making any sense since gpt4o model does support stream and temperature.
Joe Licata 0 Reputation points

2025-03-06T15:54:09.98+00:00

@santoshkc All of our assistants on east us are now receiving 404s and our quota is now -442k??? what is going on?
JAYA SHANKAR G S 3,960 Reputation points Microsoft External Staff Moderator

2025-03-07T09:41:38.6466667+00:00

Hi @Tae-Gil Noh ,

Regarding the issue our internal product team is working on, please let us know if you notice any further changes.

Thank you
santoshkc 15,355 Reputation points Microsoft External Staff Moderator

2025-03-10T07:36:04.7233333+00:00

Hi @Tae-Gil Noh,

Following up to see if you are still facing the similar issue? Do let us know if you have any further queries.
Tae-Gil Noh 30 Reputation points

2025-03-10T18:46:09.2633333+00:00
@santoshkc

Good to report back to you that it finally is working again for us. (tested 19:00 CET, sweden region)

For anyone who was tracing this issue & affected by it:

Your existing Assistants (that previously failed) may still fail -- if it was created on Azure OpenAI service and was having problem previously.

Just create a new Assistant on your OpenAI service -- rnning on the new one will no longer affected by the issue. New assistant is enough, the GPT model for the assistant can be usual pre-existing ones.

2 answers

Your answer

Ajouz, Ali 5 Reputation points

2025-02-27T13:57:25.5833333+00:00

I can confirm this problem. It impacts existing deployments
Tae-Gil Noh 30 Reputation points

2025-02-28T13:54:32.99+00:00

I think I now have more information about why some users this occurs, and why some users are okay with the Assistant.

If the Assistant is created within OpenAI resource ==> does NOT work (both new and existing)

If the Assistant is created within AIServices as "Hub" ==> does work (both new and existing)

As an example, the following picture:

AIServices "ai-useast1testhub59971" --> Assistants works okay there, no problem.

OpenAI "omq-gpt-models-us-ea..." --> Assistant does NOT work at all, as of now.

Please note that, "AIServices" resource is a "AI Hub" which has been created with "Create Project" process of AI Foundary. "OpenAI" resource is a service resource created from Azure site (just like creating any service). As I see it now, Assistants are currently working only on "AIServices" type of resource, but not with "OpenAI" resource.

To the Azure or Azure vendor personal who reads this -- this is still a terrible bug that breaks many running and pre-setup OpenAI Assistants running on Azure. Please fix this ASAP. I hope the above investigation was helpful to report / communicate the bug to AOAI dev team.
Tae-Gil Noh 30 Reputation points

2025-02-28T13:56:35.77+00:00

This is still a bug, IMHO. Please fix this ASAP!!
Marc Rousset 0 Reputation points

2025-03-03T13:53:59.13+00:00

I've been stuck on this and unable to resolve - either using OpenAI or AI Services, new and old subs. Which region are you in?
Tae-Gil Noh 30 Reputation points

2025-03-04T10:48:39.37+00:00

In my case, I was able to let it work on US east2 and Sweden central.

Maybe useful info

Both assistant and GPT created within the AIService Hub. (Linking existing resource to OpenAI didn't work).

All API calls directed to hub URL (that has microsoft cognition URLs)

The only way that I could make a hub that works with OpenAI Assistant was create a new project in AI Foundry with create new (new resource, new source group, new everything) and use that newly created hub. When I tried to connect it to previously existing resource, I've got the same errors....

However, I am not sure if this is a good workaround at all. e.g. maybe OpenAI service is running latest code, and some AI services code is not updated yet and working just for the reason -- who knows. It seems that the "preview" state of Azure OpenAI is much more volatile than "preview" state of OpenAI itself. sigh.
santoshkc 15,355 Reputation points Microsoft External Staff Moderator

2025-03-04T13:08:18.0433333+00:00

Hi @Tae-Gil Noh,

Apologies for the delay in response and thank you for reaching out and sharing these detailed observations. It looks like the issue is related to recent model updates affecting Assistants created under OpenAI resources, while those under AI Services (AI Hub) continue to function correctly. The errors around "temperature" and "reasoning_effort" suggest that certain parameters may no longer be supported for specific models.

We appreciate your insights into the difference between OpenAI resources and AI Service Hubs. If AI Hubs are working as expected, it could indicate a temporary inconsistency in how updates are being applied across different Azure environments. We recommend continuing with AI Hub-based Assistants for now while we investigate this further. Additionally, we are working on this internally to understand whether this is an expected change or a temporary issue. And please let us know if you notice any further changes or if you require additional assistance.

Thank you.
santoshkc 15,355 Reputation points Microsoft External Staff Moderator

2025-03-05T16:13:25.36+00:00

Hi @Tae-Gil Noh,

We haven’t heard from you on the last response and was just checking back to see if the given information was helpful.

Thank you.
Tae-Gil Noh 30 Reputation points

2025-03-05T18:33:54.3833333+00:00

@santoshkc your comment was not helping at all.

Azure OpenAI Assistant is still not working

Assistants, in Azure OpenAI Service, are still broken -- just tested at March 5th CET 19:20.

You didn't give us any additional information about this -- why? how? what are work arounds? when will this be fixed?

If you are an official support personal --- please be decent and provide us information like the followings

what is the current state of the problem?

is Azure OpenAI development team working for the issue already? e.g. a work ticket already out there?

when will it be fixed?

is there a solid and reliable work around? I've tried some, but as you see from the comment of others, that doesn't seem to work well for some others.

Your comment did not add anything about the issue, and basically just repeated what I have posted.

And Now you are asking : "... just checking back to see if the given information was helpful."

Sadly No, not at all:

a) We are still waiting for the issue to be fixed

b) If that takes time -- we would like to know when it will be fixed

c) If even this timeframe is not sure, then we would like to know minimally -- what is a reliable workaround for us to use Assistants APIs in a reliable way?

Thank you. We would really love to have information and/or solutions.
Tae-Gil Noh 30 Reputation points

2025-03-06T08:03:39.0933333+00:00

March 6th, 09:00 CET -- it still fails.

Posting the minimal reproducable sequence and error codes, just in case this helps Azure support to trace the issue.

Create a new Assistant on sweden-central region, on existing OpenAI service, via Assistants playground, with an existing gpt4-o (version 2024-08-06)

it doesn't matter what GPT model, version, or newly created or not -- all fails

Give "hello" to the assistant -- normally there should be a response, but here an error:

400: Unsupported value: 'stream' does not support 'true' with this model. Supported values are: 'false'. | Apim-request-id: f0e8fe72-4a73-41d9-a583-194fa264008f

Nothing works on the Assistant, for example, tried to update configuration of the Assistant, such as temperature value

Failed to update assistant ``unsupported_model: Unsupported parameter: 'temperature' is not supported with this model. Trace ID : b616aa37-2615-467f-ae8d-735a3970742cClient request ID : 3be1289d-98ed-4050-9c6e-dfcd82eff123APIM request ID : 4e89edca-ea14-4cbb-abf5-71bfc84e772b

Note that, the above error messages are not making any sense since gpt4o model does support stream and temperature.
Joe Licata 0 Reputation points

2025-03-06T15:54:09.98+00:00

@santoshkc All of our assistants on east us are now receiving 404s and our quota is now -442k??? what is going on?
JAYA SHANKAR G S 3,960 Reputation points Microsoft External Staff Moderator

2025-03-07T09:41:38.6466667+00:00

Hi @Tae-Gil Noh ,

Regarding the issue our internal product team is working on, please let us know if you notice any further changes.

Thank you
santoshkc 15,355 Reputation points Microsoft External Staff Moderator

2025-03-10T07:36:04.7233333+00:00

Hi @Tae-Gil Noh,

Following up to see if you are still facing the similar issue? Do let us know if you have any further queries.
Tae-Gil Noh 30 Reputation points

2025-03-10T18:46:09.2633333+00:00

@santoshkc

Good to report back to you that it finally is working again for us. (tested 19:00 CET, sweden region)

For anyone who was tracing this issue & affected by it:

Your existing Assistants (that previously failed) may still fail -- if it was created on Azure OpenAI service and was having problem previously.

Just create a new Assistant on your OpenAI service -- rnning on the new one will no longer affected by the issue. New assistant is enough, the GPT model for the assistant can be usual pre-existing ones.

Answer 1

Tae-Gil Noh 30

As of 10th March, the issue seems to be fixed by Azure.

You may need to create a new Assistant -- previously created Assistant with the existing problem may continue to have the problem

Answer 2

Hi @Tae-Gil Noh,

I'm glad that you were able to resolve your issue and thank you for posting your solution so that others experiencing the same thing can easily reference this! Since the Microsoft Q&A community has a policy that "The question author cannot accept their own answer. They can only accept answers by others ", I'll repost your solution in case you'd like to accept the answer.

Ask: Azure OpenAI Assistants suddenly not working properly (both Playground & API)

Solution: The issue is resolved. That As of 10th March, the issue seems to be fixed by Azure. You may need to create a new Assistant -- previously created Assistant with the existing problem may continue to have the problem

If I missed anything please let me know and I'd be happy to add it to my answer, or feel free to comment below with any additional information.

If you have any other questions, please let me know. Thank you again for your time and patience throughout this issue.

Please don’t forget to Accept Answer and Yes for "was this answer helpful" wherever the information provided helps you, this can be beneficial to other community members.

Thank You.

Saideep Anchuri 9,500 Reputation points Moderator

2025-03-12T13:52:25.51+00:00

Hi @Tae-Gil Noh,

Following up to see if the above answer was helpful. If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.

Thank You.

Share via

Azure OpenAI Assistants suddenly not working properly (both Playground & API)

2 answers

Your answer