models

Subscribe to all “models” posts via RSS or follow GitHub Changelog on Twitter to stay updated on everything we ship.

~ cd github-changelog
~/github-changelog|main git log main
showing all changes successfully

Phi-4-reasoning and Phi-4-mini-reasoning release on GitHub Models

The latest AI models from Microsoft, Phi-4-reasoning and Phi-4-mini-reasoning, are now available on GitHub Models.

Phi-4-reasoning is a model optimized for advanced reasoning in math, science, and coding. It is ideal for applications such as coding assistance, generative AI research, and knowledge-intensive problem-solving.

Phi-4-mini-reasoning is a lightweight model designed for multi-step mathematical reasoning and logic-intensive tasks, such as formal proof generation, symbolic computation, and advanced word problems. Its efficient design supports educational applications, embedded tutoring, and lightweight mobile systems.

Try, compare, and implement these models in your code for free in the playground (Phi-4-reasoning and Phi-4-mini-reasoning) or through the GitHub API.

To learn more, visit the GitHub Models documentation, and join the community discussions to share feedback and connect with other developers.

See more

What’s changing

The GPT-4o Copilot model released in preview last month and updated this month now provides code completions for all Copilot users. This model delivers higher quality suggestions and improved latency.

Getting ready

No additional action is required if you’re on the latest version of the GitHub Copilot extension for VS Code, Visual Studio, or JetBrains. If you’re on an older version of the GitHub Copilot extension, the model will roll out to you in the coming days. Updating your extension will ensure quicker access.

Retirement of GPT-3.5 Turbo based model

Over the coming days, your code completion experience will switch to GPT-4o Copilot. The GPT-3.5 Turbo based copilot-codex model, which was the previous model, will no longer be available.

Your feedback

Thank you to the tens of thousands of developers who used this new model in preview. Please continue to share your feedback directly and in the GitHub Community!

See more