How much does Copilot Fusion cost?

There is no separate charge. You are billed for each model Fusion uses at that model's standard rate, in AI credits, so Fusion High and Max can use more credits per task than Fusion Standard because they may call several models. Through November 30, 2026, paid Copilot plans get a 30% discount on model costs when using Fusion.

Which Copilot plans get Fusion High and Max?

Fusion Standard is included on every plan, including Copilot Free, Student and Pro. Fusion High and Fusion Max are included on Copilot Pro+, Max, Business and Enterprise, per GitHub's documentation on October 30, 2026.

Is Copilot Fusion as good as using Claude Opus 5.5 directly?

GitHub says that on the Artificial Analysis Coding Agent Index, Fusion Max scored within two points of Opus 5.5 at max effort, at half the cost and with 40% less time per task. That is GitHub's own announcement; check it on your repositories, especially for long tasks where one strong model keeps context better.

Quick Answer

What Is GitHub Copilot Fusion? Multi-Model Router Explained

Q: What is GitHub Copilot Fusion?

Copilot Fusion is a runtime model router and orchestrator in GitHub Copilot, made generally available at GitHub Universe on October 28, 2026. You pick Fusion in the model picker instead of a specific model; for each prompt it chooses an execution pattern (single, cascade or critique), the reasoning effort and the models, then returns one response. It is available in VS Code, Copilot CLI, the GitHub Copilot app and JetBrains IDEs.

Published: October 30, 2026

The short answer

Copilot Fusion is GitHub’s automatic model router for Copilot. Instead of choosing between models yourself, you choose “Copilot Fusion” and an effort level. For every prompt, Fusion decides whether one model is enough or whether to draft with one model and review with others, sets the reasoning effort, and returns a single answer. GitHub made it generally available on October 28, 2026 at GitHub Universe, in VS Code, Copilot CLI, the GitHub Copilot app and JetBrains IDEs.

How it works

Fusion scores each task on reasoning, code generation, debugging and tool use, then picks one of three execution patterns:

Pattern	What happens	Cost and latency
Single	One model completes the task	Same as calling one model
Cascade	An efficient model drafts; a judge accepts it or escalates to a stronger model	More calls only when needed
Critique	One model drafts; several models critique in parallel; the drafter revises	Slowest and most thorough

The choice is made per prompt, so one session can mix patterns. Fusion keeps the main conversation on one model where it can, to reuse cached tokens, and gives helper models only the context they need. It has no context window of its own: each step runs within the limits of the model running it. You can see which models ran by hovering over the effort level under the response.

Effort levels and plans (October 30, 2026)

Effort level	Priority	Plans
Fusion Standard	Balance of cost, quality and latency; one model per task	All plans, including Free, Student and Pro
Fusion High	Orchestration with moderate token savings, for multi-file changes and hard bugs	Pro+, Max, Business, Enterprise
Fusion Max	Frontier capability, still token-aware	Pro+, Max, Business, Enterprise
Fusion Fast	Fast models only; not user-selectable	VS Code inline chat

Administrators’ model policies still apply: Fusion only routes to models your plan and organisation allow. If none of its models are allowed, Fusion does not appear in the picker.

What it costs

There is no fee for Fusion itself. Each model it calls is billed at its standard rate in AI credits. A cascade or critique run can therefore cost more than a single call, while easy prompts cost less than sending everything to a frontier model. Until November 30, 2026, paid plans get 30% off model costs when using Fusion. GitHub also announced at Universe that GitHub Enterprise customers can use all Copilot features without a separate Copilot subscription, paying for the AI credits they use. That is available through GitHub Sales now, with self-serve “coming soon”. For the underlying model prices, see our current API prices.

GitHub’s headline claim: on the Artificial Analysis Coding Agent Index, Fusion Max scored within two points of Claude Opus 5.5 at max effort, at half the cost and 40% less time per task. That is a vendor-reported result.

Local models

In public preview, Fusion Standard can route to a local model: MAI Code 1.1 Flash installed on a capable Windows machine, or Qwen3.8 Flash Next served locally (for example via Ollama) in Copilot CLI. It requires turning on experimental features (“Local model auto routing” in the app, --experimental in the CLI).

Limits to know

Fusion does not start subagents; it works separately from them.
If Fusion discards a draft, file edits that draft already made are not undone automatically. Review changes before committing.
The model pool and patterns change over time, so results can shift without you changing anything.

Should you switch to it?

Use Fusion Standard as the everyday default: it is included everywhere and cheap on simple prompts. Use High or Max for multi-file refactors and stubborn bugs, and compare credit use against picking a single frontier model for a week. Keep a fixed model for work where you need reproducible behaviour, such as evaluations or regulated code. For how Copilot compares with other agents, see best AI coding agents.

Last verified: October 30, 2026.