Anthropic model formalizes Fermat's Last Theorem in Lean, 13.4M lines closing 100-theorem benchmark
littmath · x · 2026-09-05
Kevin Buzzard (Xena project) confirms that an internal Anthropic model, via the prove2.me platform, has formalized a complete proof of Fermat's Last Theorem (FLT) in Lean — the final item in Freek Wiedijk's famous list of 100 formalization challenges, closing out the 20-year-old benchmark.
- The proof follows the Darmon–Diamond–Taylor (1995) exposition of the Wiles–Taylor–Wiles argument, via the Langlands–Tunnell theorem and Ribet's level-lowering theorem.
- The repo develops Fontaine theory (flat deformations of Galois representations) and enough of Mazur's work on the Eisenstein ideal; constraints on the exponent don't matter since FLT was already formalized for odd regular primes and the smallest irregular prime is 37.
- Buzzard compiled and verified the code: over 13.4 million lines, taking 20x longer to compile than Lean's mathlib on a 96-core machine; Anthropic also provided a 500GB-RAM machine and HTML docs for navigation.
- News first leaked via a London coffee shop's Instagram, with Anthropic officially announcing an hour later.
Related event: Claude Produces First Formalized Proof of Fermat's Last Theorem in 11 Days(52 posts)→
More from Models
- Opus 5.5 tops OpenRouter spending in 20 of 29 task categories — mmurph · 2026-10-21
- Meta team builds open-source 30B LLM from scratch in months, hails 'second era' — TimDarcet · 2026-10-21
- DFlash drafter for Muse Glimmer-30B: 5 layers, blockwise 16-token spec decoding — TimDarcet · 2026-10-21
- EvoPress quantization shrinks 30B model from 56GB to 19GB at near-bf16 accuracy — TimDarcet · 2026-10-21
- Radar chart: Muse Glimmer-30B benchmarked against Gemma4-31B and Qwen3.6-27B — TimDarcet · 2026-10-21
- Privacy evals: Muse Glimmer-30B leaks less than Qwen3.6-27B on CIMemories — TimDarcet · 2026-10-21