Move off GitHub Models before the July 30 shutdown

GitHub Models is being retired for all customers on July 30, 2026. Confidence level: confirmed. GitHub says the playground, model catalog, inference API, and bring-your-own-key endpoints will stop being available after that date, including for existing active users.

What changed

GitHub's July 1 changelog gives the final shutdown date for GitHub Models. The earlier June notice only closed the product to new customers; this update affects everyone still using it.

GitHub also plans short brownouts on July 16 and July 23, 2026. During those interruptions, GitHub Models requests may temporarily return errors. Treat those brownouts as migration tests, not as the first time your app discovers the dependency.

Key takeaways

• GitHub Models closes for all customers on July 30, 2026.
• Brownouts are scheduled for July 16 and July 23.
• The retirement covers the playground, catalog, inference API, and BYOK.
• GitHub points AI model access toward Azure AI Foundry and GitHub Copilot.
• Teams should audit CI, demos, internal tools, and prompt-eval scripts for GitHub Models endpoints.

Availability and access

Existing customers can still use GitHub Models until the shutdown date, but the window is short. New customers were already blocked in June, so this is no longer a gradual adoption change. It is a hard migration deadline.

GitHub suggests Azure AI Foundry for model access and GitHub Copilot for AI workflows inside GitHub. That does not mean either is a drop-in replacement for every use case. If you used the GitHub Models API as an OpenAI-compatible endpoint, test authentication, streaming, model names, rate limits, billing, and data controls before switching traffic.

Practical LinkLoot angle

This is a dependency-risk story. GitHub Models was convenient because it bundled discovery, playground testing, inference, and BYOK in one GitHub-adjacent surface. The shutdown pushes teams back toward provider-specific APIs, Azure AI Foundry, or their own routing layer.

If you run AI features in a small product, the quickest useful move is an endpoint inventory. Search code, notebooks, docs, CI variables, and internal demo scripts for GitHub Models references. Then decide whether you need a direct provider migration or a small gateway that can swap models later.

For agent and workflow builders, pair this migration with LinkLoot's guide to AI workflow automation so model routing is not hard-coded into every prompt runner.

What to verify before you act

• Whether any production path calls a GitHub Models inference endpoint.
• Whether BYOK credentials are stored in GitHub Models settings or a separate secret manager.
• Which model names, context limits, and response formats your app depends on.
• Whether the July 16 and July 23 brownouts hit staging, production, or both.
• Whether your replacement provider keeps the same data-retention and billing assumptions.

Source check

Confirmed by: GitHub's July 1, 2026 changelog confirms the July 30 retirement, affected surfaces, all-customer scope, and brownout dates. GitHub's June 16 notice confirms the first retirement step, when GitHub Models closed to new customers.

Context: DevOps.com covered the earlier shutdown path and framed the developer migration risk. GitHub's product documentation shows the kinds of workflows GitHub Models supported, including prototyping and BYOK.