Anthropic announces Claude Fable 5.2 and Mythos 5.2. They'e already solved black hole physics problems.

Anthropic's latest models, Fable 5.2 and Mythos 5.2, represent a continued rapid development in frontier AI, following recent releases from competitors like Google, Meta, and OpenAI. The company highlights their design to address complex technical challenges, citing pre-release testing where models allegedly solved black hole physics problems, designed electronic devices, and formalized Fermat’s Last Theorem in 30 hours. These claims, if substantiated, point to advanced problem-solving capabilities.

A key distinction between the new models is their availability. Fable 5.2 is broadly accessible to all Claude users, with Anthropic stating it incorporates strong cybersecurity and biology safeguards to prevent misuse. This suggests an approach balancing advanced capability with responsible deployment for a wider audience.

In contrast, Mythos 5.2 is exclusively for verified organizations engaged in cybersecurity or biomedical research. This targeted access implies that Mythos 5.2 may possess specialized features or a different risk profile requiring controlled application within sensitive fields. Both models are reported to generate outputs 30 percent faster than their 5.1 predecessors and come with lower pricing at $8 per million input tokens and $40 per million output tokens, potentially impacting operational efficiency and cost for users.

The differentiation in access for Fable 5.2 and Mythos 5.2 indicates Anthropic is segmenting its offerings based on perceived risk and specialized application. While both are presented as powerful, the strategic choice to limit Mythos 5.2 to critical sectors suggests a recognition of its potential impact in high-stakes environments, where specific controls and verification are deemed necessary.