Anthropic released Claude Fable 5.2 and Claude Mythos 5.2, designed to capture enterprise AI spending on software engineering and research applications.
Fable 5.2 is available today; Mythos 5.2 goes only to vetted organizations. Both models claim gains in coding, reasoning, and vision capabilities, plus faster output generation.
Fable's core pitch is orchestrating developer teams and managing multi-month engineering projects. In prerelease testing, the model assisted engineers in completing software migrations previously considered infeasible and designed three working electronic devices from scratch.
Both 5.2 versions also solved two open theoretical physics problems and formalized the proof of Fermat's Last Theorem in 30 hours—capabilities Anthropic believes signal utility for advanced R&D buyers.
The gated release of Mythos to vetted customers is a deliberate commercial move. It lets Anthropic shape deployments for specific clients and negotiate higher-value contracts rather than competing on commodity pricing. Speed and reasoning improvements directly reduce engineering labor costs, the core value proposition for expensive AI systems in enterprise software shops.
