GitHub Copilot Fusion Review: Effort Levels and Pricing

Choosing an AI model has become a chore. Coding tools now offer long dropdown lists of models, and most people just pick one and hope. GitHub Copilot Fusion, generally available since October 28, 2026, is GitHub's answer: you stop picking a model and let Copilot choose, and sometimes combine, models for each task.

In this Copilot Fusion review we explain how it works, what each effort level does, how billing changes, and who it's a good fit for. We haven't run long-term tests of our own, so the performance figures here are GitHub's, and we've labelled them that way.

What is GitHub Copilot Fusion?

From the official GitHub changelog: when you select Copilot Fusion in the model picker, "you aren't selecting a specific model. You're choosing a system that routes and orchestrates models for each task." It combines two earlier GitHub projects, Project HydraFusion's multi-model orchestration and Copilot's auto model selection.

GitHub's JetBrains announcement adds that Fusion replaces the old "Auto" model option and is gradually rolling out to all Copilot users (JetBrains release notes).

Copilot Fusion effort levels: Standard, High and Max

You still make one choice: how much effort a task deserves.

• Fusion Standard: prioritises versatile models for everyday coding.
• Fusion High: prioritises capable models for challenging tasks and deeper reasoning.
• Fusion Max: prioritises the most capable models for the most demanding work.

A simple rule of thumb: Standard for small edits and explanations, High for bugs that take you more than a few minutes, Max for big multi-file changes.

How Copilot Fusion combines models

This is the clever part. GitHub describes three workflows Fusion can use:

1. Single: one model solves the task directly.
2. Cascade: an efficient model drafts a response, a quality check decides whether to accept it, and if not, the task escalates to a stronger model.
3. Critique: one model drafts, another reviews it, and the first model revises.

Cascade is how you save money: cheap models handle easy work, and expensive ones only step in when needed. Critique is how you reduce mistakes, much like asking a colleague to review your pull request.

Copilot Fusion performance: GitHub's numbers

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

Those are strong claims, and they come from GitHub, so take them as a starting point. Your results will depend on your codebase, language and how clearly you describe tasks.

Copilot Fusion pricing and billing

Billing is where you need to pay attention. GitHub says you're billed for each model Fusion uses, at each model's standard rate. So a Critique run that uses two models costs more than a Single run. For a limited time, paid Copilot plans get a 30% discount on model costs when using Fusion, and the JetBrains notes say the discount runs for the first 30 days.

GitHub's About GitHub Copilot Fusion documentation adds that there's no separate charge for Fusion itself, and that Fusion High and Fusion Max can use more than one model for a task, so they may use more AI credits than Fusion Standard.

Our advice: start on Standard, watch your usage for a week, then decide where High or Max is worth it. In JetBrains, GitHub says you can hover over an agent turn to see per-model token usage, which helps.

Where can you use Copilot Fusion?

GitHub says Fusion is available on all Copilot plans in Visual Studio Code, the GitHub Copilot app, GitHub Copilot CLI and JetBrains IDEs, with more clients coming. To try it: update your editor, open the model picker, select Copilot Fusion and choose an effort level.

Local models (experimental)

GitHub also shipped an experimental preview where Fusion can route between on-device and cloud models (local models changelog). On supported Windows devices you can download MAI Code 1.1 Flash through Windows ML. On Windows, macOS and Linux, Copilot CLI can route to models you set up yourself, such as Qwen3.8 Flash Next, through a local OpenAI-compatible server.

One important detail from GitHub: local inference "doesn't make your session offline". Fusion still makes routing decisions in the cloud, and local models may have different safety protections.

Controls for teams and admins

For companies, Fusion only uses models enabled in your enterprise and organisation policies. Admins can see which models Fusion picks from, disable specific models for Fusion without disabling them for everyone, and set Fusion at any effort level as the default.

Copilot Fusion: pros and cons

Pros

• No more guessing which model to pick.
• Cascade can cut costs on easy tasks.
• Critique adds a built-in second opinion.
• Works across VS Code, JetBrains, CLI and the Copilot app.

Cons

• Costs are less predictable, because one prompt may use several models.
• You give up some control over exactly which model touches your code.
• Local routing is still experimental.

Is Copilot Fusion good for beginners?

Yes, mostly. Beginners rarely know which model suits which task, and Fusion removes that decision. Just keep reading the code it writes. A tool that writes code faster than you can understand it doesn't make you a better programmer. If you're comparing general AI assistants for learning, see our ChatGPT vs Gemini vs Claude comparison, and for the latest OpenAI model news read our GPT-6 explainer.

Verdict

Copilot Fusion is a sensible step: most developers want good results at a fair price, not a model menu. Start on Standard, check usage, and move up only when a task earns it.

Want to put this into practice? Try our free AI Token Counter and AI API Cost Calculator right in your browser, then read Garuda Review and Recursive Flint Review on Botvana.

Related: Want AI to run tasks in your repo automatically? Read our GitHub Agentic Workflows beginner guide.

Related: For a wider pick of coding AIs (Copilot Free, Claude Code, Codex and Cursor), see Best AI for Coding 2026.

Sources and further reading

• Route and orchestrate models with GitHub Copilot Fusion (GitHub Changelog)
• GitHub Copilot in JetBrains Universe releases (GitHub Changelog)
• About GitHub Copilot Fusion (GitHub Docs)
• GitHub Copilot Fusion now supports selection of local models (GitHub Changelog)

FAQ

When did GitHub Copilot Fusion launch?

It became generally available on October 28, 2026.

Is Copilot Fusion free?

GitHub says it's available on all Copilot plans. You're billed for each model it uses at standard rates, with a 30% discount for a limited time on paid plans.

Does Copilot Fusion replace Auto?

Yes. GitHub says Fusion replaces the previous Auto model option.

Can Copilot Fusion use local models?

In an experimental preview, yes, but routing decisions are still made in the cloud.

Which effort level should I use?

Standard for everyday coding, High for harder problems, Max for the most demanding tasks.