GitHub makes Copilot Fusion generally available to pick models for you

GitHub made Copilot Fusion generally available on Oct 28. It's a new choice in the Copilot model picker, and it isn't a model. Instead of running the one model you selected, Fusion decides for each prompt which models to use and how they should work together, and you only set how much effort it puts in, according to GitHub's changelog.

Fusion combines two things GitHub already had: the multi-model orchestration from its Project HydraFusion research preview and Copilot's automatic model selection. GitHub's JetBrains release notes say it replaces the current Auto option and is rolling out gradually to all Copilot users.

There are three effort levels. Standard favours versatile models for everyday coding, High favours more capable models for harder tasks and deeper reasoning, and Max goes to the most capable models for the most demanding work. Behind each prompt, GitHub says Fusion uses one of three workflows. In Single, one model does the task. In Cascade, an efficient model drafts an answer and a quality check either accepts it or hands the task to a stronger model. In Critique, one model drafts, a second reviews, and the first revises.

The performance claim comes from GitHub, not from independent testing. It says Fusion Max scored 60 on the Artificial Analysis Coding Agent Index, within two points of Claude Code running Opus 5.5 Max, at about half the average API cost per task. GitHub also says it used about half as many tokens and took about 40% less time per task.

Billing follows the models. You're charged for every model Fusion calls, at that model's standard rate, so a Critique run that uses two models bills for both. For a limited time, paid Copilot plans get 30% off model costs when using Fusion, and the JetBrains notes put that at the first 30 days. To see how far apart model prices are, and why sending easy steps to a cheaper model can matter, see our LLM API pricing comparison.

Fusion is available on all Copilot plans in VS Code, the GitHub Copilot app, Copilot CLI and JetBrains IDEs, with more clients "coming soon." After updating, you select Copilot Fusion in the model picker and choose an effort level. Research preview users who had Auto, HydraFusion or HydraFusion Max switch to the matching levels automatically.

Administrators keep control of the model pool. Fusion only uses models enabled in enterprise and organization policies, and enterprise admins can see which models it picks from and remove a model from Fusion without disabling it for everyone. They can also make Fusion, at any effort level, the default for their users.

GitHub also put local model routing into an opt-in experimental preview the same day. At Standard, Fusion can choose between supported on-device models and cloud models in Copilot CLI, the Copilot app and VS Code. Windows gets a managed download of MAI Code 1.1 Flash. On macOS and Linux, Copilot CLI can use a model you serve yourself through a local OpenAI-compatible server, the same kind of setup covered in our OpenCode local model guide. GitHub notes that routing decisions are still made in the cloud, so a local session isn't an offline one.

Fusion moves the model choice inside one tool. If you're still deciding between Copilot, Cursor, Claude Code and Codex, our coding agent comparison covers how they differ.