OpenAI has made public a large set of AI-generated mathematical papers, announcing that the collection contains solutions to a broad range of previously unsolved questions. The release comprises 722 individual manuscripts that are organized into 372 families of related results. According to the company, the batch addresses “hundreds” of open problems across many subfields of mathematics, extending a series of recent AI-driven breakthroughs.
The papers were generated by an unreleased frontier model, and OpenAI reports that the typical proof required computational effort comparable to three hours of ChatGPT Pro usage. Earlier in September the firm had said the system resolved more than one hundred long-standing problems, but it did not disclose specific titles or publication dates until this batch appeared.
The Advisory Group on Mathematics and Artificial Intelligence, an independent panel of leading mathematicians, issued its first set of guidelines in late September. The recommendations call for immediate, peer-reviewed publication of AI-generated results, full disclosure of the model name, prompts, and compute expenditure, and avoidance of using discoveries as promotional material. The group warned that “refrain from treating the release of mathematical results as marketing vehicles to promote their models,” a practice that harms the discipline, according to AGMAI.
Mathematicians have responded with a mix of admiration and concern. While the speed and breadth of the AI-produced proofs demonstrate unprecedented capability, scholars caution that verification will take considerable time and that the community must grapple with questions of authorship, credit, and the ethical use of training data. The batch therefore adds to an ongoing debate about how AI should be integrated into formal research workflows.
OpenAI is not the only organization delivering AI-driven mathematical advances; rival lab Anthropic has also reported notable results, including work related to a Millennium Prize problem. The presence of such high-profile achievements from multiple companies underscores the rapid entry of large-scale language models into a field traditionally dominated by human insight.
The newly released manuscripts expand a growing portfolio of AI-generated mathematics that will likely shape future research norms. As the community evaluates each proof, calls for transparent reporting and responsible dissemination are expected to intensify, aiming to balance the transformative potential of machine-generated insight with the safeguards needed to preserve scientific rigor.