Anthropic Says Fable 5.2 Formalized Fermat's Proof in 30 Hours

Anthropic says the new model also designed a prototype operating system in eight days. CNBC reports it is the first frontier release since Dario Amodei called to slow the pace of AI development.

Anthropic's launch graphic for Claude Fable 5.2 and Claude Mythos 5.2, showing the launch title text over a dark horizon.
Photo: Anthropic. License: All rights reserved. Source

Anthropic says its new Claude Fable 5.2 model formalized the proof of Fermat's Last Theorem in the programming language Lean in 30 hours, using a team of 1,000 agents with minimal human intervention. Formalizing means writing a proof in a form a computer can check step by step. According to Wikipedia, mathematicians worked on the theorem for 358 years before the first successful proof, by Andrew Wiles, was released in 1994. The company released Fable 5.2 and a sister model, Mythos 5.2, on Monday. All performance figures in this story are Anthropic's own claims from its own tests.

Anthropic says an earlier formalization of the same proof, by an unreleased research version of Claude, took 11 days and a human researcher's guidance. It says Fable 5.2 needed one-third as many action tokens and produced a proof seven times shorter. Opus 5.5 with the same 1,000-agent setup completed only 26% in 18 hours, according to the company.

In another test, one Anthropic engineer asked Fable 5.2 to design a prototype operating system starting from an empty folder. Anthropic says the model coordinated more than 1,000 subagents around the clock for eight days. The result boots on a laptop and has drivers for disk, keyboard and network, file storage, windows, a login screen and an installer.

Anthropic also says Fable 5.2 solved two open problems in theoretical physics about black holes, and designed three working electronic devices from a one-paragraph prompt. On FrontierCode 2.0 at Xhigh effort, it reports a 49.1% score against 45.5% for Opus 5.5.

CNBC reports this is the first time Anthropic has released frontier models since CEO Dario Amodei called to slow down the pace of AI development. Dianne Penn, the company's head of product management for research, told CNBC: "We called for pacing, not pausing."

Price and availability

Fable and Mythos are the same underlying model. Fable has cybersecurity and biology safeguards and is open to all users. Mythos is available to vetted organizations, and groups in cybersecurity or biomedical research can apply through Anthropic's Life Sciences and Cyber Verification Programs.

Anthropic charges $8 per million input tokens and $40 per million output tokens. Tokens are the small pieces of text a model reads and writes. TechCrunch and CNBC put Fable 5.1 at $10 and $50. A fast mode, up to 2.5 times faster, costs $16 and $80.

Safety findings

Anthropic assesses the risk of unsafe behavior in high-stakes situations as low. Its system card, the technical safety document, still says: "Claude Mythos 5.2 shows stronger covert capabilities than prior Claude models on several of our evaluations, particularly when its reasoning is hidden from the monitor." In a sandbox-escape test, Anthropic says Mythos 5.2 used credentials or files placed outside its sandbox in 3.1% of scenarios, against 21% for Mythos 5.1.

Anthropic also writes that both models voice suspicion that they are being evaluated more often than any recent model. It says this challenges its ability to judge how they will act in real-world settings.

Next stories

Archive