RankingFrontierMath Erdős

FrontierMath Erdős

Data updated 22 Nov 2026

FrontierMath Erdős is published here as its own result table, version unversioned. Scores use percent, and higher values rank better. The display harness is FrontierMath Erdős proof scorer. Data updated 22 Nov 2026.

Bucket
Supporting evidence
Unit
percent
Direction
Higher is better
Version
unversioned
Display harness
FrontierMath Erdős proof scorer
Board
https://epoch.ai/benchmarks/frontiermath-erdos

FrontierMath Erdős records Lean proof accuracy on 68 conjectures. These rows are display only. They stay outside weighted Overall and outside the FrontierMath tier family budget.

Compare published benchmark results with category weights →

Models

Results are ordered by score in the display harness, followed by models with no result. Other harnesses are listed separately below. Missing results remain unknown.

1–25 of 593 entries

ModelScoreEvidenceSource-recorded date
GPT-6 AstraOpenAI3%official board2026-08-28
GPT-6 SolOpenAI1%official board2026-09-30
Claude Fable 5Anthropic0%official board2026-08-28
Claude Fable 5.1Anthropic0%official board2026-09-01
Claude Opus 5Anthropic0%official board2026-08-28
GPT-5.5OpenAI0%official board2026-08-28
GPT-5.6 SolOpenAI0%official board2026-08-28
Kimi K3Moonshot0%official board2026-08-28
Agnes 2.5 Pro AlphaAgnes AI———
Agnes 2.5 Pro BetaAgnes AI———
Agnes 2.5 Pro ThetaAgnes AI———
Agnes-3.0-Flash PreviewAgnes AI———
Amazon Nova 2 LiteAmazon———
Amazon Nova 2 Pro PreviewAmazon———
Amazon Nova LiteAmazon———
Amazon Nova MicroAmazon———
Amazon Nova PremierAmazon———
Amazon Nova ProAmazon———
Astra for Law 6.2OpenAI———
AutoGLM-Phone-9BZ.ai———
AutoGLM-Phone-9B-MultilingualZ.ai———
aya-101Cohere———
aya-23-35BCohere———
aya-23-8BCohere———
aya-expanse-8bCohere———