Skip to content

Comment on Early Evals for OpenAI O3

Comments

New SOTA's on:

SWE-Bench - 71.7

Competition Code - 2727

ARC (Semi Private Eval) - 75.7 on low, 87.5% on high compute

Frontier Math (previous SOTA was 2%) - 25% on high compute

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.