At GitHub Universe on October 28, 2026, GitHub made Copilot Fusion generally available. Instead of choosing a specific AI model for each coding task, you choose Fusion and an effort level, and Copilot decides which models to use and how they should work together.

Coffee summary

  • FACT: Copilot Fusion is a runtime model router and orchestrator, not a new model.
  • FACT: It has effort levels, including Standard, High, and Max; higher levels depend on your Copilot plan.
  • FACT: It uses three patterns: Single, Cascade, and Critique.
  • FACT: It is available in VS Code, the GitHub Copilot app, Copilot CLI, and JetBrains IDEs, and it replaces the old Auto model option.
  • CLAIM: GitHub says Fusion Max scored within two points of Opus 5.5 Max on the Artificial Analysis Coding Agent Index at half the cost and 40% less time per task.
  • OPINION: Use Standard by default and save Max for the bugs that really need it.

How Fusion works

Fusion looks at what a task needs across reasoning, code generation, debugging, and tool use. Then it picks one of three ways to run it:

  • Single: one model completes the task, just like a normal request.
  • Cascade: an efficient model drafts a solution, a judge checks it, and the work is either accepted or escalated to a stronger model.
  • Critique: one model drafts, other models review it, and the first model revises using their feedback.

GitHub's docs say the pattern is chosen per prompt, so one session can mix them. Simple tasks stay on the single pattern and do not spend extra AI credits on a review they don't need. Cascade and critique take longer because they include review calls.

Choosing an effort level

  • Fusion Standard balances cost, quality, and latency for everyday work.
  • Fusion High is meant for substantial, well-scoped tasks like a complex bug or a change across several files.
  • Fusion Max goes to the most capable models for the hardest problems.

Fusion only chooses from models your plan and your administrator allow, and GitHub says the available models may change over time. It also takes model health and availability into account, and routes at natural cache boundaries to avoid wasted cache costs.

The price angle

According to VentureBeat, users are billed for each underlying model call. GitHub's JetBrains changelog says Fusion use is discounted 30% for the first 30 days. If your team tracks AI credits closely, watch usage on High and Max during that window so the bill after the discount is not a surprise.

Who should switch

  • Solo developers: turn on Fusion Standard and stop worrying about which model is "best" this week.
  • Teams: set admin model policies first, since Fusion respects them.
  • Anyone who was in the preview: GitHub says earlier auto, HydraFusion, and HydraFusion Max selections move over to matching Fusion levels automatically.

What we still don't know

  • How the benchmark gap holds up on real, messy repositories.
  • Exact long-term pricing per effort level once the launch discount ends.

Bottom line

Copilot Fusion turns model choice into an effort dial. For most developers, that is one less decision and potentially lower cost, as long as you keep an eye on credits when you turn the dial up.

*Sources: GitHub Changelog, Copilot Fusion GA; GitHub Docs, About Copilot Fusion; GitHub Universe 2026 recap; VentureBeat.*