Anthropic announced updated models Fable 5.2 and Mythos 5.2, reporting faster inference and several claimed scientific and engineering outcomes.

Performance and developer benchmarks

According to Anthropic, both models respond up to 30% faster than their predecessors and outperform Opus 5.5 in most developer tests.

The company attributes the gains to model and inference optimizations, citing developer-oriented benchmark results shared in its announcement.

Reported scientific and engineering results

Anthropic lists three main achievements: solutions to two open problems in black hole physics, designs for three functioning electronic devices, and a formalized proof of Fermat's Last Theorem completed in 30 hours.

The firm frames these outputs as examples of the models' capabilities in reasoning, formalization and applied design within constrained testing scenarios.

Availability and access restrictions

Fable is offered broadly to users under existing subscription limits, while Mythos is provided only to vetted organizations with institutional access controls in place.

Anthropic notes that Mythos has fewer restrictions in the domains of cybersecurity and biological topics, and that access is limited to verified partners.

Announcement context

The company presented the information alongside availability details and performance claims; the initial release focused on headline results and access pathways for developers and institutions.