GitHub launched Copilot Fusion on Wednesday, a system that picks which AI models handle a coding task and how they work together, and made it generally available on every Copilot plan. Instead of choosing a model from a dropdown, developers select Copilot Fusion and let Copilot route and orchestrate the work. GitHub announced it at its Universe developer conference in San Francisco.
Fusion is available now in Visual Studio Code, the GitHub Copilot app, GitHub Copilot CLI and JetBrains IDEs, with more clients coming. It merges the multi-model orchestration from Project HydraFusion, which GitHub launched as a research preview in September, with Copilot's existing automatic model selection. Preview users who had auto, HydraFusion or HydraFusion Max selected are being moved to the matching Fusion settings.
Developers pick an effort level rather than a model. Fusion Standard is meant for everyday coding, Fusion High for harder tasks that need deeper reasoning, and Fusion Max for the most demanding work. For each task, Copilot then chooses one of three workflows: a single model solves the whole task, a cascade where an efficient model drafts and a quality check either accepts the result or escalates to a stronger model, or a critique loop where one model drafts, a second reviews, and the first revises.
“Copilot Fusion eliminates that friction, optimizing model selection without requiring developers to manage it themselves.”— Hari Pokhrel, Principal Software Engineer, AT&T
Key facts
- 01GitHub made Copilot Fusion generally available on all Copilot plans on October 28, 2026, at GitHub Universe in San Francisco.
- 02Fusion Max scored 60 on the Artificial Analysis Coding Agent Index, within two points of Claude Code with Opus 5.5 Max.
- 03GitHub says Fusion Max used about half the API cost and tokens per task and took about 40% less time.
- 04Paid Copilot plans get a limited-time 30% discount on model costs when using Fusion.
- 05By September, more than 40% of pull requests on GitHub started with an agent draft, up from 17% in March.
GitHub's headline claim is cost. On the Artificial Analysis Coding Agent Index, Fusion Max scored 60, within two points of Claude Code running Opus 5.5 Max, at about half the average API cost per task. GitHub says Fusion Max also used about half as many tokens and took about 40% less time per task. Microsoft CEO Satya Nadella described Fusion as delivering frontier-level quality 40% faster and at up to 77% lower cost.
Billing follows the models Fusion actually calls. Users pay each model's standard rate for the work it does, and for a limited time paid Copilot plans get a 30% discount on model costs when using Fusion. Enterprise and organization model policies decide which models Fusion is allowed to use, admins can switch off individual models for Fusion only, and enterprise administrators can set Fusion at any effort level as the default through managed settings.
Fusion shipped alongside a wider set of agent tools. GitHub updated the Copilot app, added Copilot projects for running agent work across several repositories in parallel, and introduced Copilot automations that can learn from their results and propose changes for developers to approve. GitHub also said GitHub Enterprise customers can now use every Copilot feature without a separate Copilot subscription, paying only for the AI credits they consume, through its sales team for now.
The launch coincided with GitHub's Octoverse 2026 data, which shows how much agent work is already flowing through the platform. By September, more than 40% of pull requests started with a draft from an agent, up from 17% in March, or 110 million in September alone. GitHub says that agent count is a floor because it cannot detect every agent. In September, 61% of Copilot users who chose their own model used more than one.
The platform has also grown underneath the agents. GitHub counted 34.3 billion commits and more than 225 million developers in the 12 months to September 30, and said it added 4 million CPU cores, 160 PB of high-speed storage and 50 Tb/s of network capacity this year. Microsoft CoreAI executive vice president Jay Parikh wrote about the scale of what comes next in a GitHub blog post.
“I think all the software we've written to date will amount to less than 1% of the software we'll write in the next five years”— Jay Parikh, Executive Vice President of CoreAI, Microsoft
GitHub is not the only company betting on routing. Replit made its intelligent model routing the default across all accounts this summer, and Ramp moved model selection into its open-source coding agent, Open SWE, in October.
Copilot Fusion is live today on all Copilot plans, with the 30% Fusion discount on model costs available to paid subscribers for a limited time.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
