⌕Is ___ SOTA?

Humanity's Last Exam

accuracy. Status active. Contamination risk low. Maintainer Scale / Epoch collection. Source.

ModelScoreDateKind
GPT-6 Astra0.54799999999999992026-09-03observation
Claude Fable 5.10.4652026-09-01observation
Gemini 3.1 Pro Preview0.46442026-02-19observation
Gemini 3.8 Flash0.445200000000000042026-09-02observation
GPT-5.4 Pro0.44322026-03-05observation
Muse Spark0.40562026-04-08observation
Gemini 3 Pro Preview0.375200000000000032025-11-18observation
GPT-5.40.36242026-03-05observation
Claude Opus 4.70.3622026-04-16observation
Claude Opus 4.60.34442026-02-05observation
GPT-5 Pro0.31642025-10-07observation
GPT-5.20.2782025-12-11observation
GPT-50.25322025-08-07observation
GPT-50.25322025-08-07observation
Claude Opus 4.50.2522025-11-24observation
Kimi K2 50.243700000000000032026-01-27observation
GPT-5.10.23682025-11-13observation
Gemini 2.5 Pro Preview 06.050.21642025-06-05observation
O30.203200000000000022025-04-16observation
GPT-5 Mini0.194400000000000022025-08-07observation
O30.1922025-04-16observation
Claude Opus 4.60.192026-02-05observation
Gemini 2.5 Pro Exp 03.250.18162025-03-25observation
O4 Mini0.18082025-04-16observation
Gemini 2.5 Pro Preview 05.060.178000000000000022025-05-06observation
O4 Mini0.14282025-04-16observation
Claude Sonnet 4.50.137200000000000022025-09-29observation
Gemini 2.5 Flash Preview 04.170.12082025-04-17observation
Claude Opus 4.10.11522025-08-05observation
Gemini 2.5 Flash Preview 05.200.109600000000000022025-05-20observation