GitHub Copilot Fusion Tops Terminal-Bench 4.1 at 71.2%, Beating Claude Code at Half the Cost
0xkarasy · x · 2026-11-18
GitHub announced at Ignite that Copilot Fusion (max) ranks #1 on Terminal-Bench 4.1 with a 71.2%±2.9% resolution rate, beating Claude Code/Opus 5.5 (65.8%) by 5 points while costing $2.0k vs $4.7k. Anthropic's Fable 5.2 ties on accuracy but costs $16.5k. Google's Gemini 4 Carbon (64.5%, $1.4k) ranks fourth; GPT-6.1 Sol is cheapest at $0.6k; Kimi K3.1 Pro and DeepSeek V4.1 Flash also chart.
Related event: GitHub Copilot Fusion Tops Terminal-Bench 4.1 at About Half the Cost(5 posts)→
More from coding & agent
- Trader wires Muse agents to execute CME futures trades, plans broker API for instant orders — CyborgTraderOFTAGI · 2026-11-18
- Qwen 3.8 27B vs Flash-Next vs DeepSeek V4.1 Flash: one-shot landing page build-off — jesdga95 · 2026-11-18
- Apple, Microsoft, FAA all hit by outages in one week — vibe-coded software meets hacker agents? — MinqiJiang · 2026-11-18
- Only 22% of security teams have a tested AI agent incident runbook, Anaconda tabletop finds — anacondainc · 2026-11-18
- Developer ships iOS app entirely from Linux — no Mac or Xcode needed — anthara_ai · 2026-11-18
- Open-source scroll-world builds Apple-style 3D scroll sites from one prompt — anthara_ai · 2026-11-18