Anthropic's Fable 5.2 formalized Fermat's Last Theorem in just 30 hours

A team of approximately 1,000 AI agents coordinated to formalize an existing mathematical proof in Lean.
Completed the task in only 30 hours with minimal human intervention.
The previous system required 11 days for a comparable task.
Used approximately three times fewer action tokens.
Produced a formal proof seven times shorter.

Anthropic releases Claude Fable 5.2 and Claude Mythos 5.2 (same underlying model), reporting that over 90% of changes merged into its core codebase are now written by Claude, with human guidance (#2).

Physics (#6): Fable 5.2 solved two open problems about black holes using holographic duality. One completed a 2003 calculation showing the transition to black hole formation is also abrupt in very small universes. The other found the exact limit of a technique used for 15+ years to rule out candidate quantum gravity theories.
Hardware (#7): from a one-paragraph prompt, Fable 5.2 designed three working electronic devices (voice-assistant hub, USB speakerphone, handheld virtual pet) with circuit boards, 3D-printed enclosures and software, with boards ready for manufacture in under a day (with Diode).
Software (#2): one engineer used Fable 5.2 to coordinate 1,000+ subagents for eight days, building a bootable prototype operating system from an empty folder (kernel, drivers, file storage, GUI, installer).
Biology (#6): Mythos 5.2 accurately modeled 28 of 34 whole adeno-associated virus (AAV) capsid structures.

Anthropic notes that during training it 'deliberately limited' the capabilities that most directly speed up AI development (accelerator kernel optimization and LLM architecture improvement), holding them at about Opus 5.5 level. HLE (with tools)=74.2%.

Nov/2026 🟠 #2 METR's independent assessment of Anthropic finds its models are improving at ~2x the previous rate, with ~80% of that acceleration attributed to AI itself. Staff self-reported speedup reached ~12x for the internal 'Model 4'. METR found 'signs of life' for fully autonomous AI contributions to the training pipeline, and no convincing bottlenecks for the next 12 months. One Anthropic researcher: 'Now only bottlenecked on ideas & compute, no longer bottlenecked on code.'

Anthropic Fable 5.2 Called Best AI Model by Far and is Pushing Engineering and Science Advancement