New to this? Read it in simple words
- GitHub’s Copilot Fusion chooses which AI models to use for each coding task.
- It is on all Copilot plans, and paid plans get 30% off for now.
- GitHub says it costs about half as much as Claude Code for a similar score.
- These numbers are GitHub’s own and have not been checked by others.
- Effort level
- A setting that tells the AI how much work, time and money to spend on an answer.
- Commit
- A saved change to the code of a project.
- Coding agent
- An AI tool that writes, tests and fixes code by itself, step by step.
Coding tools used to ask developers which AI model they wanted. GitHub now lets Copilot make that choice, prompt by prompt.
How Fusion works
In the model menu, a developer picks Copilot Fusion and an effort level: Standard, High or Max. Fusion then chooses the models for each prompt.
It can run them in three ways. Single uses one model. Cascade writes a cheap draft, checks it, and moves to a stronger model if needed. Critique has a second model review the first one’s work.
Fusion only uses models that a company’s administrator has switched on. It joins an earlier research preview, Project HydraFusion, with Copilot’s automatic model choice, GitHub says.
From GitHub’s announcements at Universe, October 28, 2026.
Who gets it, and what it costs
GitHub says Fusion is “available on all Copilot plans” in VS Code, the Copilot app, Copilot CLI and JetBrains tools. Each model it uses is charged at its normal rate. Paid plans get 30% off “for a limited time,” with no end date given.
GitHub says Fusion Max scored 60 on Artificial Analysis’s Coding Agent Index, “within two points of Claude Code with Opus 5.5 Max,” at about half the cost. It also says Fusion took about 40% less time per task. These are GitHub’s figures.
A test version can also send work to models that run on the developer’s own computer, such as MAI Code 1.1 Flash, GitHub says. The choice of model still happens in the cloud.
The rest of Universe
The Copilot app got an update, with projects, image making, and a preview of computer use on macOS and Windows. Computer use means the AI can click and type in other apps.
GitHub Enterprise now includes Copilot with no extra charge per seat, but AI use is paid for separately. Self-serve access will come in stages, GitHub says.
GitHub counted 7.4 billion commits in September, The Register and The New Stack report. It says it has added 4 million CPU cores and 160 petabytes of storage this year to keep up.
Sources
Every fact in this story comes from the sources below. Open them to check our work.
- 1
- 2
- 3
- 4Research · October 28, 2026GitHub debuts Copilot Fusion, a new model selection & orchestration system that beats Claude Code for half the API cost The New Stack
- 5Research · October 28, 2026GitHub powering up four million extra cores to stop its service going down The Register
- 6
The cost, speed and test results are GitHub’s own claims, and no outside group has checked them yet. Fusion scored a little below Claude Code, so it does not beat it on quality.
GitHub’s Copilot Fusion now picks AI models for each task. Silicon AI News, October 28, 2026. https://siliconainews.com/stories/github-copilot-fusion-universe-2026
GitHub makes Copilot Fusion available on all plans, picking AI models per task at about half the cost, by GitHub’s own figures. How we rate the mood
The mood shows how the news frames the event for the tech and AI world: as an opportunity (6 and above) or as a loss (below 5). It is not our verdict on who is right.