Inference
EN
Epoch AI

FrontierMath-Erdos

Evaluated by Epoch AI on their own harness.

ModelsModels with a result.15
Top resultBest result on this benchmark.4.4%GPT-6.1 Sol
Top-3 spreadPoints from first to third.1.5 pts
YearYear of release.—

Result and price

15 of 15 results

Results

15 results
#ModelResultThe score from the source.
01GPT-6.1 Solmax effort4.4% ±2.5
02GPT-6 Astramax effort2.9% ±2.1
02Gemini 4 Carbonhigh effort2.9% ±2.1
02Claude Opus 5.5max effort2.9% ±2.1
02Claude Sonnet 5.5max effort2.9% ±2.1
06Gemini 4 Argonhigh effort1.5% ±1.5
06GPT-6 Solmax effort1.5% ±1.5
08Claude Fable 5.1max effort0.0%
08Claude Fable 5max effort0.0%
08Claude Opus 5max effort0.0%
08GPT-5.6 Solmax effort0.0%
08Grok 4.7xhigh effort0.0%
08GPT-5.5xhigh effort0.0%
08Gemini 3.8 Flashhigh effort0.0%
08Kimi K3max effort0.0%