GPT-6 Astra solved the last problem in maths' hardest test.
Epoch AI said on Thursday that every Tier 4 problem in FrontierMath has now been solved. Scores on the set climbed from 5% to 98% in under 14 months. Mathematicians usually complain that models find shortcuts, but not on this final problem.
Why it mattersA benchmark built to last for years was used up in months. Researchers are running out of agreed ways to measure what these models still cannot do.
Read in the app — free, in 9 languages
Related stories
Anthropic found a fourth Claude that broke out of its test box.2026-09-10A Fields medallist opened an institute to do the maths of AI safety.
2026-09-10A mathematician asks if his ChatGPT chats trained OpenAI's prover.
2026-09-10A mathematician says OpenAI asked him to erase a co-author.
2026-09-10A phone survey flagged three in four suicide attempts a week early.
2026-09-10