Sonnet 5.5 is a generational leap 🤯
Sonnet 5 vs Sonnet 5.5 on quadruped physics: • Sonnet 5: Stiff, derpy stick cheetah • Sonnet 5.5: Fluid kinematics & savanna lighting • Easily rivals GPT-6 Astra & Opus 5.5 Anthropic's mid-tier just broke AI.
Claude Sonnet 5.5 ªoµ½B
ESonnet 5æè30%ȏ㑬¢ E1̍ìƂ ½èőå30%À¢i¿à\͐¦u«j EJ̕]¿ 10.3% ¨ 70.6% EüÍ2hEoÍ10h^100g[NOpus 5.5̔¼ª ¢»fÍOpusAúí̍ìƂÍSonnetցB
Sonnet 5.5 is up on RamenBench, and it absolutely EVICERATES Sonnet 5!
On High effort, it worked for 53min 54sec and used 321k tokens I don't even have to explain how much better it is than previous generation. And I noticed that starting with Opus 5.5, Claude models have
Sonnet 5.5 built a working clone of Proof, our open-source document editor, from scratch.
It did it at low effort. kieranklaassen uses this as a test for coding models. Before Sonnet 5.5, only three models he’d tested had managed it: • Fable 5 • GPT-6 Astra • Opus 5.5 Opus
You have reached the end of the archive
All of Claude Opus 5