Merge(@merge_api):Two models released in the past week now beat Claude Sonnet 5.5 (@AnthropicAI) at agentic coding. Kimi K3.1 Pro (@kimi_moonshot) passes 67.2% of Terminal-Bench 4.0, and Gemini 4 Carbon (@GoogleDeepMind), out yesterday, passes 65.2%. Kimi K3.1 Pro has the same token price as Sonnet ($2 in, $10 out per million) and costs about a third as much to run AA's Intelligence Index ($3.68 a task vs $10.02). Scores via https://t.co/TeTHwS52d4, week of Oct 26.

2026.10.29 15:00

Two models released in the past week now beat Claude Sonnet 5.5 (@AnthropicAI) at agentic coding. Kimi K3.1 Pro (@kimi_moonshot) passes 67.2% of Terminal-Bench 4.0, and Gemini 4 Carbon (@GoogleDeepMind), out yesterday, passes 65.2%. Kimi K3.1 Pro has the same token price as Sonnet ($2 in, $10 out per million) and costs about a third as much to run AA's Intelligence Index ($3.68 a task vs $10.02). Scores via week of Oct 26.