GitHub Copilot Fusion launches generally with multi-model orchestration

What is changing?

GitHub made Copilot Fusion generally available on October 28. Rather than selecting one model for each task, the system routes and orchestrates one or more models according to the need for code generation, reasoning, debugging and tool use. Users choose a Standard, High or Max effort level instead of a specific model.

Copilot Fusion can use Single, where one model answers directly; Cascade, where an efficient draft passes a quality gate and escalates when needed; and Critique, where a separate model reviews a draft before the first model revises it. It is available on all Copilot plans in VS Code, the GitHub Copilot app, Copilot CLI and JetBrains IDEs, with more clients coming later.

Enterprise administrators can limit which models Fusion uses through existing organization policies and set a default effort level. Each model is billed at its standard rate, while GitHub is offering a temporary 30% discount on model costs for paid plans. Quality, latency and cost should still be measured on real repository tasks.