Enabling Chat Completions API for GPT-5.1-mini deployment (currently only Responses API works)

Question

Hi Microsoft Support,

We have a GPT-5.1-mini deployment running on Azure OpenAI and need help enabling the Chat Completions API for it.

Current situation:

• Our deployment works fine with the Responses API (/openai/responses)
• When we send requests to the Chat Completions endpoint (/openai/deployments/{deployment-name}/chat/completions), we get a 404 Not Found error
• It looks like Azure only exposes the Responses API by default for this model

We are getting intermittent 404 "Resource not found" errors when calling our GPT-5.1-mini deployment via the Chat Completions API. The behaviour is inconsistent and we need help identifying the root cause.

Problem summary

The same deployment works from some environments and fails from others with a 404 error. In some cases it works for a period of time, then stops working without any change on our side.

What we have verified

• The API key is correct (same key works in environments that succeed).
• The resource name and endpoint URL are correct (copied directly from Azure AI Foundry > Deployment > Target URI).
• The deployment name in our request matches the deployment shown as active in the Azure portal.
• The deployment status in the portal shows as Succeeded / active.
• No configuration changes were made between the working and failing states.

Important context

We built our Chat Completions API integration based on working examples from previous deployments (GPT-4.1-mini and similar). The same URL format, headers, and request structure that worked reliably for GPT-4.1-mini now fail intermittently on GPT-5.1-mini, even though the only thing we changed was the deployment name and model.

What we need:

1. Can GPT-5.1-mini be used with the Chat Completions API on Azure? If yes, how do we enable it on our existing deployment?
2. If it requires a new deployment type or API version, please share the exact configuration (API version, endpoint path, deployment settings).
3. If Chat Completions is not supported for this model, please confirm so we can plan around it.

Answer 1

Hi Rokas Stankevicius,

Thanks for sharing your question.

If you are not able to use the Chat Completions API with your gpt-5.1-mini deployment, this is usually related to how the endpoint is being used or how the model is accessed.

You can check the following points step by step.

First, understand the default behavior In Azure AI Foundry, the portal often shows the Responses API by default. This does not mean Chat Completions is not supported. You can still use Chat Completions if the model supports it. [en.ittrip.xyz]

Second, use the correct endpoint format For Chat Completions, the endpoint must include the chat path and your deployment name.

Example format

• https://your-resource.openai.azure.com/openai/deployments/your-deployment-name/chat/completions

Make sure you are not using the responses endpoint by mistake. [bing.com]

Third, verify the deployment name Use the exact deployment name created in the portal, not the model's name.

For example,

• correct: your deployment name
• incorrect: gpt-5.1-mini (if this is not the deployment name)

Using the wrong name can result in errors or API not working.

Fourth, check API version Make sure you are using a supported API version for chat completions.

Some common versions

• 2024-08-01-preview
• 2024-06-01

Using an incorrect version may cause the request to fail or not recognize the API.

Fifth, confirm model capability Not all models behave the same way. Some newer models may prefer or default to the Responses API.

In general

• GPT-5 base and mini models support chat-based interactions
• But certain variants may require correct endpoint usage to work properly [github.com]

Sixth, validate request format Make sure your request body follows the chat format:

• include messages array
• include role and content
• avoid unsupported parameters

Chat models expect structured conversation input, not plain text prompts.

Useful reference

• Work with chat completion models
• Example using Chat Completions API

In short: the issue usually comes from using the wrong endpoint or deployment name, not because Chat Completions is disabled.

I Hope this helps. Do let me know if you have any further queries.

Thankyou!

Answer 2

GPT-5.1-mini can be used with the Chat Completions API in Azure OpenAI in Foundry, but it must be called through the correct endpoint shape and API version for chat completions.

Key points based on the available information:

1. Supported Chat Completions endpoint shape

For chat completions in the Foundry Models REST API preview, the documented endpoint is:

POST https://{endpoint}/openai/deployments/{deployment-id}/chat/completions?api-version=2025-04-01-preview

Required elements:

• endpoint: https://{your-resource-name}.openai.azure.com
• deployment-id: the name of the GPT-5.1-mini deployment
• api-version: 2025-04-01-preview (for the preview chat completions API)

Headers (one of the following auth methods):

• Authorization: Bearer {Azure_OpenAI_Auth_Token} with a token obtained for https://cognitiveservices.azure.com
• or api-key: {your-azure-openai-key}

If the current calls are using a different api-version, a different base URL, or a path that does not match the above, a 404 Not Found is expected. Ensure:

• The base URL is exactly https://{your-resource-name}.openai.azure.com (no /openai/v1/ suffix for this preview REST shape).
• The path is /openai/deployments/{deployment-id}/chat/completions.
• The query string includes ?api-version=2025-04-01-preview.

2. Relationship to Responses API

The Responses API and Chat Completions API are two different surfaces over the same models:

• Responses API: typically called via /openai/responses (often through the Foundry SDKs like AIProjectClient and get_openai_client().responses.create(...)).
• Chat Completions API: called via /openai/deployments/{deployment-id}/chat/completions with the preview API version.

The same deployment (for example, a GPT-5.1-mini deployment) can be addressed via:

• responses.create(model=<deployment-name>, ...) using the Foundry Responses client, or
• POST /openai/deployments/{deployment-name}/chat/completions?api-version=2025-04-01-preview using the Chat Completions REST API.

No special “deployment type” toggle is documented for enabling chat completions separately from responses; the critical requirement is using the correct endpoint and API version.

3. Checklist to resolve the 404

Validate the following against the current implementation:

1. Endpoint
   • Must be: https://{your-resource-name}.openai.azure.com
   • Do not append /openai/v1/ when using the Foundry Models REST preview shape shown above.
2. Path
   • Must be: /openai/deployments/{deployment-name}/chat/completions
   • {deployment-name} must exactly match the deployment name configured in Azure OpenAI / Foundry.
3. API version
   • Use: api-version=2025-04-01-preview for chat completions in the Foundry Models REST API preview.
4. Authentication
   • Use either:
     • Authorization: Bearer {token} where the token is obtained via:

              az account get-access-token --resource https://cognitiveservices.azure.com
       

       and the scope https://cognitiveservices.azure.com/.default, or
     • api-key: {your-azure-openai-api-key}.

If any of these elements are incorrect (for example, using an older api-version, a different base URL, or a mismatched deployment name), the service can return 404 Not Found even though the deployment works via the Responses API.

4. When to assume chat completions are not available

From the provided documentation, chat completions are exposed via the preview endpoint above. There is no indication in the context that GPT-5.1-mini is excluded from chat completions. If, after aligning the endpoint, path, and api-version as described, the deployment still returns 404 specifically for chat completions while Responses continues to work, this would indicate a service-side or configuration issue that requires support investigation rather than a model limitation.

In that case, capture:

• Full request URL (with api-version),
• Deployment name,
• Resource name/region,
• Exact HTTP status and body,

and provide them to Azure support for further analysis.

---

References:

• Azure OpenAI in Microsoft Foundry Models REST API preview reference
• Azure OpenAI in Azure AI Foundry Models REST API preview reference
• Work with chat completions models
• Work with chat completions models (programming-language-dotnet)
• Use model router for Microsoft Foundry
• Azure OpenAI in Microsoft Foundry Models REST API reference