← All notes
AI

GitHub Copilot Fusion adds new models and wider availability

DV
Daan VisserNovember 18, 2026 · 2 min read

GitHub announced today that Copilot Fusion is being expanded with two new AI models, broader surface coverage, and benchmark results posted on two public Terminal‑Bench leaderboards.

What’s new?

The rollout adds a ‘Fusion‑Lite’ model aimed at lower compute usage and a ‘Fusion‑Pro’ model that delivers higher‑quality code suggestions. Both run on the same transformer backbone but are tuned for speed versus accuracy respectively. They are immediately usable in GitHub Copilot for VS Code, JetBrains IDEs, and the new Copilot CLI.

Availability and impact

Beyond the existing IDE integrations, Fusion is now exposed in the GitHub Pull‑Request reviewer and in GitHub Codespaces. This means developers get AI assistance throughout the entire workflow, from local editors to cloud‑based development environments.

Benchmark results

GitHub posted scores on two Terminal‑Bench leaderboards: one measures response time in interactive terminal sessions, the other evaluates the quality of generated shell scripts. Fusion‑Pro is 23 % faster than the previous model and scores 15 % higher on script quality. The blog post contains full methodology details.