GitHub Copilot Fusion: Terminal-Bench 4.1 Result, Wider Availability and New Models

GitHub CopilotView original changelog

Copilot Fusion now appears on the public Terminal-Bench 4.1 leaderboards, where Fusion Max reached a 71.2% resolution rate. GitHub says that is 5.4 points ahead of Opus 5.5 and Fable 5.2, at 88% lower cost than Fable 5.2 and 57% lower than Opus 5.5. Fusion is also now selectable on github.com, Visual Studio and the GitHub Mobile beta apps, and three models joined its tiers.

Key Takeaways

  • Fusion Max resolved 71.2% of Terminal-Bench 4.1 tasks, which GitHub says is 5.4 points ahead of Opus 5.5 and Fable 5.2.
  • GitHub reports 88% lower cost than Fable 5.2 and 57% lower than Opus 5.5 on that benchmark.
  • Results sit on public leaderboards hosted by Stanford, Harbor and the Laude Institute rather than only on GitHub's own site.
  • Fusion is now selectable on github.com and Visual Studio, plus the GitHub Mobile beta apps.
  • Production mobile support is described as coming shortly, so most mobile users cannot use it yet.
  • Three models joined tiers: Claude Sonnet 5.5 in Max, GPT-6.1 Terra in High and MAI-Code-2 in Standard.

Copilot Fusion Gets Benchmarks, Reach and Models

Copilot Fusion lets a developer pick an effort level while Copilot selects and coordinates models for each task. GitHub shipped three updates to it: a public benchmark result, broader availability, and new models in each tier.

Terminal-Bench 4.1 Results

Copilot Fusion is now listed on the public Terminal-Bench 4.1 leaderboards hosted by Stanford, Harbor and the Laude Institute. Fusion Max reached a 71.2% resolution rate. GitHub says that is 5.4 percentage points ahead of Opus 5.5 and Fable 5.2. It also reports that Fusion Max cost 88% less than Fable 5.2 and 57% less than Opus 5.5 on the benchmark.

Where Fusion Can Be Used

Fusion can now be chosen in the model picker on the latest versions of github.com and Visual Studio, and in the GitHub Mobile beta apps for iOS and Android. GitHub says support in the production mobile apps is coming shortly. In each place, the developer sets an effort level and Copilot handles model selection and coordination.

New Models in Fusion Tiers

  • Claude Sonnet 5.5 was added to Copilot Fusion Max.
  • GPT-6.1 Terra was added to Copilot Fusion High.
  • MAI-Code-2 was added to Copilot Fusion Standard.

These additions change what each tier draws on, so results and cost at each effort level can shift.