GitHub Copilot Fusion tops Terminal-Bench 4.1 at 71.2% for ~$2k, beating Opus and GPT
kimmonismus · x · 2026-11-18
A GitHub Universe stage slide shows GitHub Copilot Fusion claiming #1 on Terminal-Bench 4.1 with 71.2%±2.9% accuracy at roughly $2.0k cost (3.3B tokens), versus Claude Opus 5.5 at 63.8% ($47k), Claude Fable 5.2 at 63.8% ($16.5k), Gemini 4 Carbon at 64.5% ($1.4k), Sonnet 5.5 at 63.8% ($7.2k), and OpenAI GPT-6.1 Sol near 60.2% ($334k). Fusion was released Oct 28, 2026.
Related event: GitHub Copilot Fusion Tops Terminal Bench 4.1 at Half the Cost(3 posts)→
More from coding & agent
- Microsoft announces MAI Code 2 for GitHub Copilot — testingcatalog · 2026-11-18
- Tencent Hunyuan releases Accelerate Bench: 32 CPU tasks to grade coding agents' optimization — _akhaliq · 2026-11-18
- GPT-6.2 Astra Ultra put to the test: turn a dream-house PDF into a Blender model and playable game — BLCNYY · 2026-11-18
- /prune-tests: a Claude command to hunt down useless tests — teropa · 2026-11-18
- SpaceXSI staffs its accounting org with 5 Grok Bots, each owning a job end to end — elonmusk · 2026-11-18
- Microsoft launches Copilot Workspace for team collaboration at Ignite — donasarkar · 2026-11-18