✦
GitHub has made Copilot Fusion generally available, a new system that chooses which AI models handle each coding task instead of asking the developer to pick one. The launch took place on 28 October at GitHub Universe 2026, alongside a new Copilot app and Copilot projects.
Fusion is a model router and orchestrator. Rather than selecting a model from a list, a developer chooses Copilot Fusion and an effort level, Standard, High or Max, and the system then picks one or more models and a working pattern for each prompt. It can run models one after another, in sequence, or have them check each other's work. The idea is to match the effort to the job, using a light touch for simple edits and heavier resources for hard problems.
The system is available on all Copilot plans in Visual Studio Code, the GitHub Copilot app, the Copilot command-line tool and JetBrains IDEs, with more clients promised. It builds on Project HydraFusion, a research preview that began on 4 September, combined with Copilot's automatic model choice. GitHub says Fusion Max scored 60 on the Artificial Analysis Coding Agent Index, within two points of Claude Code with Opus 5.5 Max, at about half the average cost per task and about 40 percent less time per task.
Fusion is billed per model at each model's standard rate, with no separate fee for the routing itself. Users on paid Copilot plans receive a 30 percent discount on model costs when using Fusion through 30 November 2026. Enterprise customers can also make Fusion the default model for their teams and set a default effort level.
The launch reflects a shift in coding tools, from a single model doing everything to a system that picks the right model for each step. For developers, the promise is less time spent choosing models and more time spent on the work itself, although the cost will depend on which models Fusion calls and how often it calls them.
Source: Unite.AI
