GitHub announced Copilot Fusion at its Universe developer conference in San Francisco on October 28, 2026, and made it generally available the same day. Fusion is not a new AI model — it is a system inside the Copilot model picker that decides, for each coding task, which models to use and how to combine them, so developers no longer have to pick a model themselves.

The change is about cost as much as capability. Developers who select Fusion choose only an effort level: Standard for everyday coding, High for harder problems that need deeper reasoning, and Max for the most demanding work. Fusion merges two earlier pieces of GitHub technology — the multi-model orchestration of Project HydraFusion, released as a research preview in September, and Copilot's automatic model selection. GitHub says there is no extra orchestration fee: each model used is billed at its standard rate, and paid Copilot plans get a 30% discount on Fusion model costs through November 30, 2026.

GitHub's headline benchmark claim is striking, though it is the company's own figure, not an independent test. The company says Fusion Max scored 60 on the Artificial Analysis Coding Agent Index, within two points of Claude Code running Anthropic's Opus 5.5 Max, while using roughly half the average API cost and tokens per task and completing tasks about 40% faster. GitHub's COO told reporters that internal teams and early enterprise adopters are seeing roughly 50–60% lower costs on complex work, with some cases reaching 77%.

Fusion is available on all Copilot plans in Visual Studio Code, the GitHub Copilot app, the Copilot command-line tool, and JetBrains IDEs. The launch came alongside other Universe announcements, including the GitHub Copilot app, Copilot Projects, and Copilot code review Max, plus a change for GitHub Enterprise customers, who can now use every Copilot feature without a separate subscription and pay only for the AI credits they consume. GitHub's annual Octoverse report, published the same day, said agents wrote the first draft of more than 40% of pull requests by September, up from 17% in March — a sign of how fast agent-written code is spreading.

For developers, the practical question is whether Fusion's savings hold up on their own codebases. If GitHub's numbers generalize, Fusion could make frontier-level coding help affordable for smaller teams; if not, it is still a useful simplification — one picker option instead of a growing menu of models. Either way, it signals a shift in the coding-assistant race: from "which model should I use?" to "let the system decide."

Sources