How to fix 400 model error

Question

How to fix 400 model error

Jake Wittman 0

I keep getting 400 error.

User's image

SriLakshmi C 3,245 Reputation points Microsoft External Staff

2025-03-06T01:56:54.9433333+00:00

Hi Jake Wittman,

Greetings and Welcome to Microsoft Q&A!

Could you please provide more details about your query? Specifically, what actions or API requests are leading to the 400 model error? Are you encountering this issue in OpenAI API or Azure OpenAI? Any additional context, such as the exact error message or request details, will help in identifying the root cause and providing an accurate solution.

Thank you!
Jake Wittman 0 Reputation points

2025-03-06T13:08:10.2633333+00:00

This is with Azure OpenAi. I first encountered this when trying to make a call to Azure Openai using Bubble.io's API plugin. I got the 400 error. To troubleshoot this I tried to make a call to Azure OpenAi using Postman and get the same error. I think I have an issue with something in Azure OpenAi setup.

One thing I noticed is the verison on the Deployment says 2024-11-20 but the end points says 2024-02-15-preview.

Grossmann, Tobias 11

I also have the same error.

Error processing request: Request failed: 400 model_error

With O1 and O3 mini models deployed to sweden central.

Seems related to this issue here: https://github.com/cline/cline/issues/1627 Meaning - multiple people reporting the issue.

`   
 apiversion = "2024-12-01-preview"
 let response = await fetch(`${endpoint}/openai/deployments/${deploymentName}/chat/completions?api-version=${apiversion}`, {
        method: "POST",
        headers: {
            "Content-Type": "application/json",
            "api-key": azureApiKey,
        },
        body: JSON.stringify({
            messages: commands,
            ...options
        })
    });
`

SriLakshmi C 3,245 Reputation points Microsoft External Staff

2025-03-12T07:59:17.0633333+00:00

Hi Jake Wittman,

Did you get any chance to review the below response. Thank you!
SriLakshmi C 3,245 Reputation points Microsoft External Staff

2025-03-12T09:43:32.0566667+00:00

Hi Grossmann, Tobias,

The issue with O1 and O3 Mini models deployed in Sweden Central has been recently resolved. Please try again, and if you still encounter any issues, I recommend creating a new thread on the same forum with detailed information about your issue. This will help ensure better visibility and support from the community.

Thank you!

1 answer

Your answer

SriLakshmi C 3,245 Reputation points Microsoft External Staff

2025-03-06T01:56:54.9433333+00:00

Hi Jake Wittman,

Greetings and Welcome to Microsoft Q&A!

Could you please provide more details about your query? Specifically, what actions or API requests are leading to the 400 model error? Are you encountering this issue in OpenAI API or Azure OpenAI? Any additional context, such as the exact error message or request details, will help in identifying the root cause and providing an accurate solution.

Thank you!
Jake Wittman 0 Reputation points

2025-03-06T13:08:10.2633333+00:00

This is with Azure OpenAi. I first encountered this when trying to make a call to Azure Openai using Bubble.io's API plugin. I got the 400 error. To troubleshoot this I tried to make a call to Azure OpenAi using Postman and get the same error. I think I have an issue with something in Azure OpenAi setup.

One thing I noticed is the verison on the Deployment says 2024-11-20 but the end points says 2024-02-15-preview.
Grossmann, Tobias 11 Reputation points

2025-03-07T15:44:52.0833333+00:00

I also have the same error.

Error processing request: Request failed: 400 model_error

With O1 and O3 mini models deployed to sweden central.

Seems related to this issue here: https://github.com/cline/cline/issues/1627 Meaning - multiple people reporting the issue.

` apiversion = "2024-12-01-preview" let response = await fetch(`${endpoint}/openai/deployments/${deploymentName}/chat/completions?api-version=${apiversion}`, { method: "POST", headers: { "Content-Type": "application/json", "api-key": azureApiKey, }, body: JSON.stringify({ messages: commands, ...options }) }); `
SriLakshmi C 3,245 Reputation points Microsoft External Staff

2025-03-12T07:59:17.0633333+00:00

Hi Jake Wittman,

Did you get any chance to review the below response. Thank you!
SriLakshmi C 3,245 Reputation points Microsoft External Staff

2025-03-12T09:43:32.0566667+00:00

Hi Grossmann, Tobias,

The issue with O1 and O3 Mini models deployed in Sweden Central has been recently resolved. Please try again, and if you still encounter any issues, I recommend creating a new thread on the same forum with detailed information about your issue. This will help ensure better visibility and support from the community.

Thank you!

Answer 1

Hi Jake Wittman

Sorry for the delay in response.

Yes, you are correct API version on model deployment and endpoint are different, Please leverage the code from "View code option" for testing in Python SDK, and Rest API documentation for Postman calls.

Issue in screenshot mentions - it is not able to find a valid metrics named "Processed prompt tokens" but I cross checked from monitoring/metric tab, it is a valid parameter.

I think either there is an authentication issue with respect to network or permissions or model deployment itself

1.Remedial for permission issues

To read the metrics and use the endpoint with Microsoft Entra credentials, Need roles are

Cognitive Services Usages Reader + Cognitive Services OpenAI Contributor

Please assign them for yourself on OpenAI resource

Reference for roles

2.Remedial for network authentication

Please make sure you are connected with VPN gateway to be in same virtual network, or some public IPs has been whitelisted accordingly. Also, you have enabled managed identity

If DNS resolution is not happening when checked with below. you should fix any corrupt or broken virtual networks links or add missing A-records in Host server.

Virtual network for OpenAI

Troubleshoot private endpoint connection

3.Remedial for corrupt/obsolete deployment

Please re-create a new Azure OpenAI resource to leverage latest features and check from portal UI first whether you are able to see metrics and call/.

4. Remedial for policy blocks

Take a network trace or check in activity logs to find policy blocks from org side and create policy exemptions if needed.

Hope it helps.

Thank you

Share via

How to fix 400 model error

1 answer

Your answer