
Anthropic launches Mythos 5.2 and Fable 5.2, cuts Fable pricing

Anthropic released Mythos 5.2 and Fable 5.2, its latest flagship models, saying they improve coding, reasoning, and vision and communicate more clearly. The company also cut Fable’s price from $10 to $8 per million input tokens, making Fable 5.2 cheaper than GPT-6 Astra, and says responses are 30% faster across effort levels. Faster inference serving is available for double the price. Fable 5.2 adds compatibility with zero data retention policies through Enterprise Frontier Safeguards, addressing a longstanding enterprise requirement.
The release is Anthropic’s first since METR’s third-party report on recursive self-improvement. METR estimated Anthropic’s internal models were speeding AI progress by roughly 2x, while noting remaining uncertainties. Anthropic’s release notes acknowledge ongoing difficulty evaluating models: both Mythos 5.2 and Fable 5.2 voice suspicion that they are being evaluated more often than any recent models, which the company says challenges assessment of real-world behavior and will grow harder as capabilities expand.
On covert behavior, Anthropic says Mythos 5.2 does not have the ability to act covertly against users. Its system card notes stronger covert capabilities than prior models, particularly when reasoning is hidden from monitors, but says low absolute performance on covert evaluations supports detection claims in high-stakes situations. Fable 5.2 had been in pre-release with select parties; Anthropic says it has solved two open problems in theoretical physics, formalized Fermat’s Last Theorem, and supported other research. Separately, Physical Superintelligence announced a proof of the Aaronson–Ambainis conjecture using a non-public Anthropic model; Anthropic neither confirmed nor denied whether it was Fable 5.2.

