GitHub Copilot Fusion is generally available: Copilot routes the models
GitHub Copilot Fusion is now generally available. Instead of picking a model per task, developers select Fusion and an effort level, and Copilot decides which models run and how they cooperate. Administrators can restrict which models Fusion may use and make it the default. It is available in VS Code, the GitHub Copilot app, Copilot CLI and JetBrains IDEs.
Key points
- Pick Copilot Fusion in the model picker, then set the effort level: Standard, High or Max
- Fusion Standard targets everyday coding; Fusion Max targets the most demanding, reasoning-heavy tasks
- Workflows: Single (one model), Cascade (cheap draft plus quality check) and Critique (draft, review, revise)
- Admins can limit which models Fusion uses via enterprise and organization model policies
- Billing follows each underlying model's standard rate; paid plans get a 30% model cost discount for a limited time
- Research preview users' auto, HydraFusion and HydraFusion Max choices move to matching Fusion levels
GitHub has made Copilot Fusion generally available, merging Project HydraFusion's multi-model orchestration with the model selection behind Copilot's auto mode. You pick Fusion and an effort tier instead of a model, and Copilot decides per prompt which models run and how they work together, from a single direct answer to a cheap draft that gets escalated or reviewed and revised.
Administrators can limit Fusion to approved models and make it the default. Fusion Max scored 60 on the Artificial Analysis Coding Agent Index, close to Claude Code with Opus 5.5 Max, at about half the cost per task. Billing follows each underlying model's rate, with a temporary 30% discount for paid plans, and the feature runs in VS Code, the GitHub Copilot app, Copilot CLI and JetBrains IDEs.
Real-world use
- Teams that do not want to switch models for every task can pick an effort level and let Copilot handle the rest
- Raising the effort to High or Max for a hard refactor or architecture analysis that needs stronger reasoning
- Organizations that must control spend can restrict Fusion to an approved model list via policy
Why it matters
You no longer have to pick a model for each step: choosing an effort level now sets the cost and quality trade-off. Administrators can also centrally decide which models Copilot may use on their behalf.