Skip to content
WeChat · AI Era (Xinzhiyuan)·Indexed by this site · Originally published· AI updates →

Anthropic releases Claude Fable 5.2 and Claude Mythos 5.2, claiming 1000 agents formally proved Fermat's Last Theorem in 30 hours

最强Fable 5.2震撼发布!费马大定理30小时形式化,全面碾压Opus 5.5

AI summary

Anthropic releases Claude Fable 5.2 and Claude Mythos 5.2, claiming Fable 5.2 completed a formal proof of Fermat's Last Theorem with 1000 agents within 30 hours. Fable 5.2 scores 65 on AA Intelligence Index v5 and 81% on ARC-AGI 3, reduces the safety false-block rate by about 55%, and the company says its specific AI R&D capabilities are suppressed to the Opus 5.5 level; Mythos 5.2 has an attempt rate of 3.1% across 1800 sandbox jailbreak scenarios (Mythos 5.0 was 39%), and finds 10.2% more network vulnerabilities than the previous strongest model.

Read originalmp.weixin.qq.com