OpenAI’s release o1 lands in GitHub Copilot Chat and Models

OpenAI has shipped the production version of its o1 reasoning model, and GitHub is making it available immediately in Copilot Chat and on GitHub Models. The release improves on the earlier o1-preview iteration with better performance on complex, multi-step work; OpenAI reports a 44% improvement on Codeforces’ competitive coding benchmark versus the preview model.

The updated model is available now in Copilot Chat across Copilot Pro, Business, and Enterprise plans—but not on the free Copilot tier. In Visual Studio Code and on GitHub’s web interface, the model picker now shows o1 (Preview), where the parenthetical “Preview” refers to the model picker itself, not to o1.

Trying o1 in Copilot Chat

Subscribers can use o1 for the usual Copilot workflows: explaining code, debugging, refactoring, modernization, and writing tests. The model also pulls context from the current workspace and from GitHub repositories, so it can reason over your actual project state rather than only isolated snippets.

o1 usage is included in existing paid Copilot subscriptions, with a rate limit of 10 messages per 12 hours. If you’re on a Copilot Business or Enterprise plan, an organization administrator must first enable access to o1 models before the model appears in the picker.

A GIF showing a developer selecting the OpenAI’s o1 model in GitHub Copilot to power Copilot Chat within the integrated development environment (IDE) VS Code.

o1 on GitHub Models

Developers who want to experiment before committing to Copilot can try o1 through GitHub Models, which offers a playground and an API for prototyping AI applications. From the playground, you can compare o1 side by side with models from Mistral, Cohere, Microsoft, or Meta, then grab the generated code samples to start building against the API.

A GIF showing a sample prompt with OpenAI’s o1 model in GitHub Models, a platform that enables developers to experiment with leading AI models, compare their performance, and build them into applications. The sample prompt asks “How do you prove the Pythagorean theorem geometrically.” The AI model then offers an answer that displays its reasoning capabilities.

Keeping multiple models available reflects GitHub’s broader stance that developer choice matters as much as raw model capability—whether that means picking o1 for hard reasoning tasks or another provider’s model for other workloads. The o1 rollout is part of a continued push to surface newer OpenAI models inside GitHub’s tooling as soon as they ship.