REST API Support for Claude and Gemini Models

[Mohammed-Basheer]

Select Topic Area

Question

Body

The GitHub Models REST API currently only supports OpenAI models, while Copilot Chat already provides access to Claude and Gemini models. Is there a plan to expose Claude and Gemini through the /inference/chat/completions REST API endpoint so the API experience matches what’s available in Copilot Chat?

[frazrajpoot01]

Hey! The reason for this disconnect actually comes down to the backend infrastructure.

Copilot Chat routes requests dynamically, which allows it to connect to Anthropic and Google's servers to give you access to Claude and Gemini in your IDE.

However, the GitHub Models REST API is heavily powered by Azure AI under the hood. Azure natively hosts models from OpenAI, Meta, Mistral, and DeepSeek, which is why those are all readily available on the /inference endpoint. Since Azure does not host Claude or Gemini, they aren't in the GitHub Models API catalog.

GitHub hasn't shared a public roadmap or timeline for if/when they will bridge this gap for the REST API. For now, if you need programmatic API access to Claude or Gemini, you'll still have to get keys and hit Anthropic or Google's APIs directly (or use AWS/GCP).

Hope that clarifies the current setup!

[M677871]

Not today, as far as the public docs show.
The REST API for GitHub Models and Copilot Chat do not expose the exact same model set, so Claude and Gemini being available in Copilot Chat does not automatically mean they are available through /inference/chat/completions.

Unless GitHub announces support for them in the Models API, the safe assumption is that they are not supported there yet.

[hardik121121]

Great question, and the existing answers explain the infrastructure side well. Just want to add a bit more practical context:

The core issue is that GitHub Models and Copilot Chat are actually two separate systems that happen to share a UI surface.
GitHub Models'/inference/chat/completions endpoint is essentially a proxy to Azure AI Foundry, which is why it follows Azure's model catalog. Copilot Chat, on the other hand, has direct partnerships/routing to Anthropic and Google that bypass that Azure layer entirely — hence the mismatch.

As of now, there's no public commitment from GitHub to bring Claude/Gemini into the Models REST API. The GitHub roadmap discussions haven't surfaced anything concrete on this.

If you need programmatic access today, your practical options are:

• Anthropic API directly — api.anthropic.com with an API key from console.anthropic.com
• Google AI Studio / Vertex AI for Gemini
• LiteLLM if you want a unified OpenAI-compatible interface across providers without changing your code structure much

One thing worth watching: GitHub has been expanding the Models catalog over time, so it's not impossible this changes
— but I wouldn't hold a project dependency on it without an official announcement.