GitHub Copilot Fusion Routes Each Task Across Multiple AI Models

GitHub Copilot Fusion is now generally available. Instead of choosing a specific model for each task, developers select Copilot Fusion in the model picker, and Copilot routes and orchestrates models for each prompt. The feature combines Project HydraFusion’s multi-model orchestration with the fine-tuned model selection of Copilot auto model selection. GitHub says it has improved Project HydraFusion’s performance since that project launched.

GitHub reports that Copilot Fusion Max scored 60 in the Artificial Analysis Coding Agent Index, within two points of Claude Code with Opus 5.5 Max. It reached that result at about half the average API cost per task, using about half as many tokens and taking about 40% less time per task.

Developers choose an effort level: Fusion Standard for everyday coding, Fusion High for challenging tasks, or Fusion Max for the most demanding work. Copilot then picks a workflow, from a single model solving the task directly to a cascade or critique process that adds a quality check or a review and revision step.

Why it matters

Choosing models task by task takes effort and can waste time or money. Copilot Fusion moves that decision into the tool, and GitHub’s benchmark figures suggest it can approach a leading coding agent at lower cost. Administrators keep control, which matters for teams with strict policies.

What to do

Select Copilot Fusion in the model picker and compare Fusion Standard with Fusion High on your real tasks. If you manage an organization, review your enterprise and organization model policies, since Copilot Fusion only uses enabled models and can be set as the default. Confirm pricing, plan availability, and usage limits on GitHub’s site, because the release notes do not cover them.