Skip to main content

GitHub Copilot Fusion: One Click, Multiple AI Models

GitHub Copilot Fusion: Let the System Pick the Model

Choosing an AI model can feel like picking a restaurant for a group of indecisive friends—everyone has an opinion, and you just want to eat. GitHub's new Copilot Fusion aims to end that debate. Announced at GitHub Universe 2026, it's an intelligent orchestration layer that automatically routes coding tasks to the best AI model from OpenAI, Anthropic, Google, and others—no manual selection required.

Image

How It Works

Developers simply select Copilot Fusion in the model selector. The system then assesses task complexity and decides whether to use a single model, a Cascade mode (where one model builds on another's output), or a Critique mode (where models review each other's work). It also matches the task to one of three effort levels: Standard (efficiency and cost), High (enhanced reasoning), or Max (maximum capability). This prevents simple tasks from burning through expensive compute.

Performance and Cost

On the Artificial Analysis Coding Agent Index benchmark, Copilot Fusion Max scored 60 points—just 2 points shy of Claude Code running Opus5.5Max. But here's the kicker: average API cost was about half, token consumption was cut in half, and tasks finished roughly 40% faster. That's a rare win-win in the AI coding space.

Copilot Fusion is now generally available across all GitHub Copilot plans, including Visual Studio Code, CLI, the app, and JetBrains IDEs. There's no extra orchestration fee—you only pay for actual model calls. And through November 30, 2026, model fees are 30% off. Enterprise administrators can control which models are available and set Fusion as the default experience. The system even supports experimental local model routing, like calling Microsoft MAI Code1.1Flash on Windows devices.

Microsoft's Broader Play

Microsoft also launched Copilot harness, developed by the GitHub team. It already powers products like Copilot Studio and VS Code, and is available to developers through Microsoft Foundry. The move signals a broader trend: multi-model intelligent orchestration is becoming a key direction for AI coding tools, promising a better balance between performance and cost.

Key Points

  • Automatic model selection: Copilot Fusion routes tasks to the best model from multiple providers without manual choice.
  • Three execution modes: Single model, Cascade, or Critique, matched to Standard, High, or Max effort levels.
  • Strong benchmark results: 60 points on the Coding Agent Index, with half the cost and 40% faster completion compared to top competitors.
  • Wide availability: Works across VS Code, CLI, GitHub app, and JetBrains IDEs; no orchestration fee, 30% off model fees until Nov 30, 2026.
  • Enterprise controls: Admins can manage model access and set Fusion as default; local model routing is experimental but supported.