GitHub makes Copilot Fusion generally available

Copilot Fusion routes and orchestrates models for each task, so developers do not have to select a specific model every time.[1] GitHub says it combines Project HydraFusion’s multi-model orchestration with Copilot’s automatic model selection.[1] Standard prioritizes versatile models for everyday coding; High prioritizes capable models for challenging tasks and deeper reasoning; Max prioritizes the most capable models for demanding tasks and complex reasoning.[1] Administrators can control which models Fusion uses and whether it is the default.[1]

GitHub says Fusion users are billed for each model it uses at that model’s standard rate, with a 30% discount on Fusion model costs for paid Copilot plan users for a limited time.[1] Fusion is available on all Copilot plans in VS Code, the Copilot app, Copilot CLI and JetBrains IDEs.[1] It can solve a task with a single-model workflow, or use Cascade: an efficient model drafts a response that can be accepted or escalated to a stronger model.[1] In Critique, one model drafts, another reviews, and the first revises the response.[1]

Artificial Analysis measured an average cost of $5.59 per task for Fusion, compared with $11.42 for Claude Code with Opus 5.5.[2] It measured 7.2 million tokens per Fusion task, about half the tokens used by Claude Code with Opus 5.5.[2] GitHub says Fusion took about 40% less time per task.[1] These figures describe measured tasks; teams evaluating costs should also account for Fusion’s per-model billing and the limited-time discount terms.[2][1]

Copilot Projects are in public preview in the Copilot app, bringing project context, agents and results together.[3] Users can set a project goal and add context such as repositories, design mockups, Slack threads or meeting notes; GitHub says Copilot can run as many agents as a project needs and return results such as pull requests for review.[3] The app can draw on connected tools including Slack and Figma and supports generating and editing images through natural-language chat.[3] Computer Use is in public preview on macOS and Windows; GitHub says it asks for approval before controlling an app, and organizations can disable it through managed settings.[3]

GitHub says intelligent routing for local models is in experimental preview.[4] Fusion can choose between supported on-device and cloud models in Copilot CLI, the Copilot app and VS Code.[4] GitHub also says Fusion Standard improved token efficiency across AT&T by automatically selecting the right model for each task.[1] The announcement describes that result for AT&T; it does not establish the same outcome for other organizations.[1]