1,000 Claude agents redid the Fermat proof in 28.7 hours

Anthropic says 1,000 agents on an early version of Claude Fable 5.2 reproduced its formal proof of Fermat’s Last Theorem in 28.7 hours, without the Prove2Me platform or a researcher steering. It’s one toy run with a head start, since the agents got a sketch of the classical route and proved lemmas from the September formalization.
The September run took a researcher and dozens of agents 11 days. Anthropic itself says the result doesn’t replace tools like Prove2Me, because agents tend to give up on hard problems and misjudge their difficulty.
Each agent ran in a sandbox with no internet, and a classifier blocked none of the team’s tool calls. Shortly after the run, Anthropic added a second monitor that reads the agents’ chain of thought.
Up next
Meta’s Muse Ember matches Haiku at about a third the cost
At $0.14 per task against $0.41 for Claude Haiku 5.5, Meta’s Muse Ember matches Haiku’s score by Meta’s own estimate, while charging the same per-token rates. Is your go-to small model right now Haiku or something else?
Related stories
- Fermat’s Last Theorem gets a machine-checked proof
- Fable 5.2 matches Opus 5.5 on FrontierCode at half the cost
- Claude Fable 5.2 designed three working gadgets with Diode
- METR estimates Anthropic now progresses about 2x faster
- Anthropic’s FlashProve proves Llama 4 Maverick at 110 tok/s
- Claude vouches for Team and Enterprise users at plugin login
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

