GitHub made Copilot Fusion generally available, routing work across models instead of making you pick one

GitHub made Copilot Fusion generally available on October 28, announcing it at GitHub Universe 2026. Instead of choosing a model for each task, you choose Fusion in the model picker and Copilot decides which models to use and how they work together. You only pick an effort level.

Fusion combines the multi-model orchestration of Project HydraFusion with Copilot's fine-tuned auto model selection. The three levels are Standard, for versatile everyday coding, High, for challenging tasks and deeper reasoning, and Max, for the most demanding work. For each prompt it runs one of three workflows: Single, where one model solves the task; Cascade, where an efficient model drafts and a quality check accepts the result or escalates to a stronger model; and Critique, where one model drafts, another reviews, and the first revises.

GitHub's own benchmark claim: on the Artificial Analysis Coding Agent Index, Copilot Fusion Max scored 60, within two points of Claude Code with Opus 5.5 Max, at about half the average API cost per task, about half the tokens and about 40% less time per task. That comparison comes from GitHub, not an independent test of our own.

Billing and availability: you are billed for each model Fusion uses at that model's standard rate, and Unite.AI reports paid Copilot plans get a limited-time 30% discount on model costs when using it. It is available on all Copilot plans in Visual Studio Code, the GitHub Copilot app, Copilot CLI and JetBrains IDEs. People who used auto, HydraFusion or HydraFusion Max in the research preview move to the matching Fusion level automatically.

Administrators keep control. Fusion only uses models enabled in enterprise and organization policies, admins can disable a model for Fusion without disabling it for everyone, and they can set Fusion as the default through enterprise managed settings. AT&T's Hari Pokhrel said Standard improved token efficiency by selecting the right model for the task, since developers often pick a powerful model and forget to switch back.

What to try: switch to Fusion Standard for a few days of normal work and compare spend and quality with your usual model, then try Max on a hard task. The practical question is whether the savings GitHub reports hold up on your own codebase.

Building with AI? See this week's hot AI repos and how to use each one.