OpenAI has released 722 research manuscripts detailing solutions to hundreds of long-standing open problems in mathematics, all generated by an unreleased frontier artificial intelligence model. Published in a GitHub repository alongside reasoning summaries and compute estimates, the collection spans 372 distinct result families. According to OpenAI, generating the average result required roughly three hours of continuous thinking using its high-end ChatGPT Pro tier.
The release follows recommendations issued in late September by the Advisory Group on Mathematics and Artificial Intelligence (AGMAI), an independent panel of elite mathematicians formed to encourage responsible communication of AI proofs. AGMAI urged artificial intelligence developers to publish findings promptly through established academic channels, disclose model names, prompts, and compute costs, and avoid using mathematical breakthroughs primarily as marketing devices. OpenAI stated that it is implementing protocols for citations and paper revisions while evaluating other community-hosted repository options.
The new papers add to a rapidly expanding volume of mathematical discoveries produced by OpenAI and competing laboratories like Anthropic, including work addressing a Millennium Prize problem. However, the swift entry of corporate AI labs into pure mathematics has sparked intense debate over research ethics, academic standards, and how tech companies credit human mathematicians whose work underpins these automated solutions.