How to Use GitHub Copilot Fusion to Pick Models for You

I switch models in Copilot all day. A cheap one for a rename, a strong one for a tricky bug, then I forget to switch back and pay for the strong one on a typo fix. GitHub Copilot Fusion is GitHub's answer to this. It became generally available on October 28, and it picks the models for you.

What is new in GitHub Copilot Fusion

Fusion is a new entry in the Copilot model picker. You choose Fusion and an effort level, and Copilot decides which model handles each prompt. The official changelog lists three levels:

1. Standard for everyday coding. It picks one model per task.
2. High for harder, well scoped tasks.
3. Max for the most complex work.

On High and Max, Fusion can also combine models. In a "cascade", a cheaper model writes a draft and a quality check accepts it or sends it to a stronger model. In a "critique", one model writes, a second one reviews, and the first one fixes. It works in VS Code, JetBrains IDEs, the Copilot CLI and the GitHub Copilot app, on all Copilot plans.

Who it helps

A freelance developer. You pay for your own AI usage, so every wasted call on a big model is your money. Standard as your default keeps small tasks on cheaper models without you thinking about it.

An engineering team lead. You get questions like "which model should I use?" and surprise bills at the end of the month. Admins can set Fusion as the default and control which models it may use, so the team gets one sane default instead of ten personal habits.

How people are using it so far

It is only a few days old, and I did not find real user stories yet. The only story I found is the one GitHub published itself: an AT&T engineer says the team no longer has to switch back to cheaper models for normal work. GitHub also says Fusion Max scored within two points of Claude Code with Opus 5.5 Max on the Artificial Analysis Coding Agent Index, at about half the cost per task. That is GitHub's own number, so test it on your code.

How to try GitHub Copilot Fusion

1. Update your editor extension or the Copilot CLI.
2. Open the model picker and choose Copilot Fusion.
3. Start with Standard for a day of normal work.
4. Switch to High or Max only for one hard task, and compare the result and the cost.

A good test prompt for High:

Find why the order total is wrong when a discount and free shipping apply together.
Write a failing test first, then fix it.
Do not change the public API of OrderService.
At the end, list every file you changed.

The honest part

Fusion is not free magic. You pay the normal rate for every model it uses, and High and Max can use more than one model per task, so one task can cost more than before. Paid plans get 30% off model costs with Fusion until November 30, 2026. If Fusion throws away a draft, file edits from that draft are not undone, so review the diff before you commit. I found no EU specific limits, but if your company restricts Copilot to data resident models, Fusion only uses those.

If you want more practical AI tips like this, subscribe to my newsletter. One short entry a week, straight to your inbox.

Sources

1. Route and orchestrate models with GitHub Copilot Fusion (GitHub Changelog)
2. Copilot Fusion (GitHub Docs)
3. GitHub Copilot Fusion now supports selection of local models (GitHub Changelog)
4. GitHub debuts Copilot Fusion (The New Stack)
5. Copilot Fusion feedback discussion (GitHub Community)