FrontierMath saturated in two years: new GPT solves Tier 4 100%, 9/49 open problems down
Afinetheorem · x · 2026-10-03
Afinetheorem notes that mathematicians' skeptical statements about AI on FrontierMath Tier 3 look shocking less than two years later: Tier 3 has been saturated, the harder Tier 4 was solved 100% by the new GPT, and 9 of 49 "Open Problems" — even harder — are already solved. He adds that bottlenecks remain, but progress at even 10% of this pace applied to biology, pharma, construction, green energy and education is the goal — failing to achieve it would be history's greatest anti-Progress fumble.
Related event: FrontierMath Saturated in Two Years as New GPT Aces Tier 4(2 posts)→
More from Models
- vllm-sr open-sources Decision 2.0 models in six sizes from 0.6B to 27B — victormustar · 2026-10-04
- davidad: labs are already prioritizing functional infinite context, but consumer memory is held back — davidad · 2026-10-04
- ChatGPT voice's constant 'mmhmms' annoy users who just want to work — MicahBerkley · 2026-10-03
- Reddit User Details Gemini's Denials: Forgets Context, Calls Real News Fake, Ends Chat — CaptainDucki · 2026-10-03
- OpenAI's new reasoning models can have chats human-reviewed even if you opted out — JeremyNguyenPhD · 2026-10-03
- Gemini refuses to recognize a Google Home automation, but only in Android Auto — magnumforce2006 · 2026-10-03