Back to rankings

GitHub Copilot Fusion: Multi-Model Routing Now Generally Available

GitHub CopilotView original changelog

GitHub made Copilot Fusion generally available, letting developers pick Fusion instead of a specific model while Copilot routes and orchestrates models for each task. It merges the HydraFusion research preview鈥檚 multi-model orchestration with Copilot auto model selection and offers Standard, High and Max effort levels. GitHub reports that Fusion Max scored within two points of Claude Code with Opus 5.5 Max on the Artificial Analysis Coding Agent Index at roughly half the cost per task. Paid plans get a limited-time 30% discount on model costs when using Fusion.

Key Takeaways

  • Fusion replaces manual model picking, routing each task to the right model so developers no longer switch models themselves.
  • Fusion Max scored 60 on the Artificial Analysis Coding Agent Index, within two points of Claude Code with Opus 5.5 Max, according to GitHub.
  • GitHub claims about half the cost and tokens and roughly 40% less time per task compared with that top-scoring setup.
  • Three workflows (Single, Cascade, Critique) let Fusion trade speed, cost and quality per task.
  • Standard, High and Max effort levels give developers control over how capable the model pool is.
  • A limited-time 30% discount on paid plans and admin controls for allowed models and defaults make adoption easier for teams.

What Copilot Fusion Is

GitHub made Copilot Fusion generally available across all Copilot plans. Instead of choosing a specific model for every task, developers select Fusion in the model picker and Copilot decides which models to use and how to combine them. Fusion brings together the multi-model orchestration first shown in the HydraFusion research preview and Copilot's auto model selection into one experience.

Effort Levels

Fusion offers three effort levels. Standard uses versatile models suited to everyday coding. High uses more capable models for challenging tasks that need deeper reasoning. Max uses the most capable models for demanding, complex work. Developers who selected auto, HydraFusion or HydraFusion Max during the research preview are moved automatically to the matching Fusion effort level.

Three Execution Patterns

For each task, Fusion picks one of three workflows. In the Single pattern, one model solves the task directly. In the Cascade pattern, an efficient model drafts a response and a quality check either accepts it or escalates to a stronger model. In the Critique pattern, one model drafts, a second model reviews, and the original model revises based on that feedback.

Performance and Cost

GitHub reports that on the Artificial Analysis Coding Agent Index, Fusion Max scored 60, within two points of Claude Code with Opus 5.5 Max. According to GitHub, it reached that result at about half the average API cost per task, with about half as many tokens and around 40% less time per task. AT&T reported better token efficiency and less wasted spend because developers no longer switch models by hand.

Billing and Admin Controls

Usage is billed per model that Fusion uses, at each model's standard rate. For a limited time, users on paid Copilot plans receive a 30% discount on model costs when using Fusion. Fusion only uses models enabled in enterprise and organization model policies, and administrators can disable specific models for Fusion without disabling them for everyone. Enterprise managed settings can also make Fusion, at any effort level, the default model for Copilot users.

Availability

Fusion is available in Visual Studio Code, the GitHub Copilot app, GitHub Copilot CLI and JetBrains IDEs, with more clients planned. To try it, update the client, open the model picker, choose Copilot Fusion and select an effort level.