OpenAI · AI-Generated Mathematics

OpenAI Math Manuscripts

A repository, published 6 October, of 719 mathematics manuscripts in 372 families written by an unreleased internal OpenAI model, which the company pointed at about 4,000 open research problems after its existing math evaluations saturated. Each result used on average about three hours of ChatGPT Pro thinking compute, and about 42% of headline results (300 of 719) come with Lean formalizations; the README warns that unformalized results may contain errors. Reasoning summaries are included for ten families, among them the irrationality exponent of pi and the Mezard-Parisi spin-glass formula. On 7 October OpenAI withdrew three Hodge-conjecture-related papers over a sign error and revised 14 others. Most results are not peer reviewed. Apache-2.0, 13,745 stars.