Developer choice comes to Copilot

GitHub is opening up Copilot to multiple AI models, letting developers pick the best tool for each task instead of being locked into a single provider. Anthropic's Claude 3.5 Sonnet, Google's Gemini 1.5 Pro, and OpenAI's o1-preview and o1-mini are all being integrated into the platform.

Rollout starts in Copilot Chat, with OpenAI's o1 models available immediately, Claude 3.5 Sonnet arriving progressively over the next week, and Gemini 1.5 Pro following in the coming weeks. GitHub plans to extend multi-model support across Copilot Workspace, multi-file editing, code review, security autofix, and the CLI. Individual developers can choose their preferred models directly in VS Code or on GitHub.com, while organizations and enterprises retain full administrative control over which models their teams can access.

AI-native workflows across the platform

Workspace, Code Review, and Autofix convergence

Copilot Workspace is being integrated directly into pull requests, allowing developers to refine, validate, and land AI-generated code suggestions from Copilot code review, Copilot Autofix, and third-party Copilot Extensions. The goal is to shorten the path from initial suggestion to merged pull request.

A screenshot of a GitHub Copilot Workspace editor for a pull request showing Copilot assisting a developer by suggesting code to fix an authentication error. On the left, there is a list of files in the pull request. On the right, a conversation between the developer and Copilot shows that Copilot has identified the syntax error in the code and is suggesting a fix for it, with a natural language explanation. The middle panel shows the failed build and the Typescript code, with Copilot suggested code inserted. At the top of the screen there is a green button labeled ‘Review and commit’ and a button showing three total suggestions.

GitHub Spark: micro apps from natural language

GitHub Spark is a new technical preview aimed at making software creation accessible to a much broader audience. It's designed for small, personal, and fun applications rather than enterprise productivity tools—the kind of experimental projects people often don't build because the setup overhead isn't worth it.

Spark is powered by natural language prompts and includes features like live history, previews, and direct code editing. The intent is to let anyone take a small idea and turn it into a working micro app without wrestling with infrastructure.

A screenshot of GitHub Spark, an experimental application by GitHub Next, on a mobile phone. The application enables people to build software with natural language prompts. The top of the application has a text box asking, ‘What do you want to create?’, a ‘Generate variants?’ checkbox, and a drop-down list with ‘GPT-4o’ selected. Below that, there is a list of recent sparks titled Carter Boys Allowance Tracker, Find my City, Spark News, and Tic-Tac-Taco.

Platform-wide productivity improvements

Beyond the AI announcements, GitHub is rolling out a range of enhancements across the developer platform. The company says these updates target the building, releasing, scaling, and securing of software, aiming to reduce friction throughout the entire development lifecycle.

Copilot Workspace's pull request integration, combined with the multi-model rollout, positions Copilot as a more central part of the development loop—not just a code completion tool, but a system that participates in code review, security fixes, and the triage of incoming suggestions.