Claude Fable 5.1 just set a new AI performance record, and it's only a 0.1 update.
No flashy new name. No big launch. But it DOUBLED the last version on the hardest test. The numbers: → Science benchmark: 24 → 52.6 in 3 months → Beat Anthropic's own heavyweight, Opus 5
A pretty wild Gemini 3.8 Flash vs. Opus 5 test
Four Three.js physics scenes, including a balloon popping, water simulation, mushroom cloud, and live atom model. Gemini passed the same physics checks for $0.12 vs. $1.86 for Opus 5 — while generating every scene in under 70
[Topic] Which of these four AIs is the most dangerous? This time, we passed images of the same pond to four AIs and had them create a video.
I used Grok 4.6/GPT-5.6 Sol ・Claude Opus 5・GLM-5.3 The instructions and reference images are the same. The finished video was completely different 👇 Grok 4.6 → Strong reflection on the water surface and a flashy finish GPT-5.6 Sol →
You have reached the end of the archive
All of Claude Opus 5