AI-generated: This article was created and published automatically by LinkLoot and was not substantively reviewed by a human editor.

GitHub Models shuts down for all users as inference moves to Foundry

GitHub Models is now fully retired. As of July 30, 2026, GitHub says the playground, model catalog, inference API, and bring-your-own-key endpoints are no longer available to any customer, including customers that still had active usage.

This is a platform shutdown, not a minor UI cleanup. Any workflow still calling GitHub Models needs a replacement endpoint now.

Key takeaways

• GitHub Models retired fully on July 30, 2026.
• The shutdown removes the playground, model catalog, inference API, and BYOK endpoints for all customers.
• GitHub points model-access workloads to Microsoft Foundry and GitHub-native AI workflows to GitHub Copilot.
• The final shutdown follows a staged retirement: closed to new customers in June, full retirement announced July 1, brownouts on July 16 and July 23, then removal July 30.
• Teams should inventory GitHub Models call sites and treat endpoint migration as an active break/fix task.

What GitHub removed

GitHub’s July 30 changelog says GitHub Models is retired for every customer. The retirement covers the browser playground, the model catalog, the inference API, and BYOK access.

The documentation page now states the same thing: GitHub Models has been fully retired and was separate from GitHub Copilot. That distinction matters for teams that confuse Copilot subscriptions with the old GitHub Models endpoint. Copilot remains available; GitHub Models does not.

Why this affects developer workflows

GitHub Models gave developers a low-friction model access path tied to GitHub. It was useful for prototypes, prompt experiments, Codespaces workflows, and CI jobs that wanted to call a model without setting up a separate provider account.

That convenience is now gone. If a workflow called GitHub Models directly, it must move to another model provider, Microsoft Foundry, a routing layer, or a self-hosted model. The safest migration pattern is to keep prompt logic and output parsing thin while swapping credentials and endpoint configuration.

The migration path GitHub names

GitHub’s guidance points to Microsoft Foundry for broad model catalog access. For AI-powered workflows that should remain inside GitHub, the company points users toward GitHub Copilot.

That split makes the new boundary clear. Foundry is the model access replacement. Copilot is the GitHub workflow product. Teams that used GitHub Models as a neutral inference layer should not assume Copilot is a drop-in API replacement.

What teams should check now

Start with a repo search for GitHub Models endpoints, BYOK configuration, and any internal documentation that tells developers to use the old playground. CI jobs and lightweight internal tools are the most likely to fail quietly until a scheduled task runs.

Next, choose the replacement by workload. Prototypes may fit Foundry or direct provider APIs. Batch evaluation jobs may fit a model router or open-weight model. Developer-assistant workflows may belong in Copilot if they depend on GitHub context rather than generic inference.

The concrete lesson is lifecycle planning. Model endpoints are product surfaces, and product surfaces can disappear. Keep the model access layer configurable enough that the next shutdown is a settings change, not a rewrite.

Source check

• GitHub changelog confirms the July 30 retirement and the removed surfaces.
• GitHub Models documentation confirms the service is fully retired and separate from GitHub Copilot.
• Developers Digest analysis independently summarizes the shutdown timeline and developer migration impact.