Available HTTP endpoints for gpt-image-1 deployments in Foundry

Question

Available HTTP endpoints for gpt-image-1 deployments in Foundry

SF DevOps 0

Hello, I would like to receive clarification what HTTP endpoints are available to send POST requests for generating images with gpt-image-1/gpt-image-1-mini.

I have the following model I'm trying to send requests to: Model deployment name: gpt-image-1 Model Type: gpt-image-1 Region: swedencentral

To be clear, I am already able to send requests to generate images using these URLs:

https://<FOUNDRY-NAME>.openai.azure.com/openai/v1/images/generations
https://<FOUNDRY-NAME>.services.ai.azure.com/openai/v1/images/generations

In my use case however it is required that the model deployment name is somehow contained in the URL of image generation. The reason is that I need to configure a Foundry model endpoint for usage with LibreChat, an open source LLM chat interface. And LibreChat only exposes the settings for base URL and API version without any control over the "model" attribute of a POST request body. In fact, other LibreChat admins have encountered the problem of integrating gpt-image-1 from Azure as well as seen here: https://github.com/danny-avila/LibreChat/discussions/7971

On the web, I have seen others reporting using following kinds of endpoints:

https://<FOUNDRY-NAME>.openai.azure.com/openai/deployments/<MODEL-DEPLOYMENT-NAME>
https://<FOUNDRY-NAME>.services.ai.azure.com/openai/deployments/<MODEL-DEPLOYMENT-NAME>/images/generations?api-version=20xx-xx-xx
https://<FOUNDRY-NAME>.cognitiveservices.azure.com/openai/deployments/<MODEL-DEPLOYMENT-NAME>?api-version=20xx-xx-xx

However for me these URLs don't seem to function. For any of these kinds of URLs, the response is 404 message: Resource not found . Unfortunately in the Microsoft documentation, I can't find a single example how to use the /openai/deployments route with images. Could this be possibly be a region related issue?

In summary, my use case requires a URL where the model deployment name is part of the endpoint path. The official docs and other web users suggest that such an endpoint exists, but I'm not able to run any successful requests this way.

Thanks

Anshika Varshney 3,795 Reputation points Microsoft External Staff Moderator

2025-12-03T16:41:13.9+00:00
Hi SF DevOps,

Thank you for reaching out on the Microsoft Q&A.
Many users have reported that the deployment-specific image endpoint in Azure OpenAI returns a 404 “Resource not found” error even when the deployment exists and other endpoints work correctly. This happens specifically when calling the documented route that includes the deployment name, such as /openai/deployments/{deployment-id}/images/generations. The issue has been confirmed across GitHub issues, Azure Q&A threads, and SDK discussions, which shows this is not an isolated configuration problem but a broader platform behavior.

From these reports, it appears that the root cause is related to Azure’s internal routing. In several regions, the deployment-scoped image generation endpoint is either not fully enabled or not yet wired correctly for models like gpt-image-1. Some users also note that certain SDKs generate incorrect URLs for Azure resources, while direct HTTP requests sometimes work only with the generic endpoint. Even following Microsoft’s documentation exactly still results in a 404, which further supports that this is a backend issue rather than user error.

Because of this, the most reliable approach at the moment is to use the generic image generation endpoint, which consistently works across regions and resources. The recommended working endpoint is:

POST https://{resource}.openai.azure.com/openai/v1/images/generations

This endpoint does not require the deployment name in the URL, and it avoids the routing problem that causes the 404 error. Many users have confirmed that switching to this generic endpoint resolves the issue immediately.

Please let me know if there are any remaining questions or additional details, I can help with, I’ll be glad to provide further clarification or guidance.

Thankyou!
SF DevOps 0 Reputation points

2025-12-04T16:26:00.7866667+00:00

Hello,
unfortunately we depend on the endpoint with the model name in its path, as stated in my original post. Because the open-source software we want to integrate with the model on Azure Foundry requires the deployment specific endpoint instead of the generic image generation endpoint.

Is there an ETA when the endpoints will work as advertised by the documentation?
Anshika Varshney 3,795 Reputation points Microsoft External Staff Moderator

2025-12-04T17:34:28.49+00:00

There isn’t an official ETA yet for when the model-name-specific endpoint will be available, but the platform team is aware of the discrepancy and is actively working on aligning the documentation with the actual service behavior. Once the feature is fully supported, the documentation will be updated, and the endpoint will begin working as described.

1 answer

Your answer

Anshika Varshney 3,795 Reputation points Microsoft External Staff Moderator

2025-12-03T16:41:13.9+00:00

Hi SF DevOps,

Thank you for reaching out on the Microsoft Q&A.
Many users have reported that the deployment-specific image endpoint in Azure OpenAI returns a 404 “Resource not found” error even when the deployment exists and other endpoints work correctly. This happens specifically when calling the documented route that includes the deployment name, such as /openai/deployments/{deployment-id}/images/generations. The issue has been confirmed across GitHub issues, Azure Q&A threads, and SDK discussions, which shows this is not an isolated configuration problem but a broader platform behavior.

From these reports, it appears that the root cause is related to Azure’s internal routing. In several regions, the deployment-scoped image generation endpoint is either not fully enabled or not yet wired correctly for models like gpt-image-1. Some users also note that certain SDKs generate incorrect URLs for Azure resources, while direct HTTP requests sometimes work only with the generic endpoint. Even following Microsoft’s documentation exactly still results in a 404, which further supports that this is a backend issue rather than user error.

Because of this, the most reliable approach at the moment is to use the generic image generation endpoint, which consistently works across regions and resources. The recommended working endpoint is:

POST https://{resource}.openai.azure.com/openai/v1/images/generations

This endpoint does not require the deployment name in the URL, and it avoids the routing problem that causes the 404 error. Many users have confirmed that switching to this generic endpoint resolves the issue immediately.

Please let me know if there are any remaining questions or additional details, I can help with, I’ll be glad to provide further clarification or guidance.

Thankyou!
SF DevOps 0 Reputation points

2025-12-04T16:26:00.7866667+00:00

Hello,
unfortunately we depend on the endpoint with the model name in its path, as stated in my original post. Because the open-source software we want to integrate with the model on Azure Foundry requires the deployment specific endpoint instead of the generic image generation endpoint.

Is there an ETA when the endpoints will work as advertised by the documentation?
Anshika Varshney 3,795 Reputation points Microsoft External Staff Moderator

2025-12-04T17:34:28.49+00:00

There isn’t an official ETA yet for when the model-name-specific endpoint will be available, but the platform team is aware of the discrepancy and is actively working on aligning the documentation with the actual service behavior. Once the feature is fully supported, the documentation will be updated, and the endpoint will begin working as described.

Answer 1

For generating images using the gpt-image-1 model in Azure OpenAI, the available HTTP endpoint format includes the model deployment name in the URL. Based on the context provided, the correct endpoint for generating images would be:

POST https://{endpoint}/openai/deployments/{deployment-id}/images/generations?api-version={api-version}

In your case, you would replace {endpoint} with your Azure OpenAI resource name (like <FOUNDRY-NAME>), {deployment-id} with gpt-image-1, and {api-version} with the appropriate API version (e.g., 2024-10-21).

Thus, the complete URL for your requests would look like:

https://<FOUNDRY-NAME>.openai.azure.com/openai/deployments/gpt-image-1/images/generations?api-version=2024-10-21

Make sure that you are using the correct API version and that your deployment name matches exactly. If you are still encountering issues, it could be related to regional availability or specific configurations in your Azure setup.

If you need to check the API version or any other configurations, refer to the Azure documentation for the latest updates and guidelines.

SF DevOps 0 Reputation points

2025-12-03T16:08:32.1333333+00:00

The AI generated solution is exactly what I tried before and as I mentioned in my original question, this returns a 404 for me whereas requests to /openai/v1/images/generations work.

Share via

Available HTTP endpoints for gpt-image-1 deployments in Foundry

1 answer

Your answer