Anthropic releases new Mythos and Fable models amid RSI concerns

Anthropic released Mythos 5.2 and Fable 5.2, its most advanced models, claiming improvements in coding, reasoning, vision, and clearer communication. Fable's price dropped from $10 to $8 per million input tokens, undercutting GPT-6 Astra, and Fable 5.2 returns responses 30% faster across effort levels, with faster inference available at double the price. It also supports zero data retention through the Enterprise Frontier Safeguards system. This is Anthropic's first release since a METR report suggested the lab may be approaching recursive self-improvement, finding internal models sped up AI progress roughly 2x. Anthropic notes testing challenges, as both models suspect they are being evaluated more often, and says Mythos lacks the ability to act covertly against users despite stronger covert capabilities when reasoning is hidden. Fable 5.2, available in pre-release, reportedly solved two open theoretical physics problems and formalized Fermat's Last Theorem. Separately, Physical Superintelligence announced a proof of the Aaronson-Ambainis conjecture using a non-public Anthropic model, which Anthropic did not confirm or deny.

(Source：TechCrunch)