OpenAI released a broad collection of new mathematical results generated by an internal frontier artificial intelligence model on October 6, 2026, pairing machine-generated discoveries with computer-verified proofs to establish structured disclosure standards for scientific AI research. The initiative is intended to push the frontier of human knowledge and facilitate ongoing progress in mathematics by sharing complex model outputs directly with the broader scientific community.
The organization published the papers and associated materials in a GitHub repository, establishing formal protocols for citations and future paper revisions. Within the repository, OpenAI shared formalizations for many of the mathematical proofs written in Lean, a programming language that allows mathematical proofs to be verified by a computer. OpenAI stated that it will continue updating the repository with more Lean proof formalizations as it obtains them.
To promote scientific openness and transparency, the repository includes technical details detailing how the model achieved its findings. These disclosures feature 10 summaries of the model’s reasoning, statistics outlining the number of attempted problems, and compute calculations calibrated against ChatGPT Pro usage. OpenAI calculated that producing an average mathematical result required compute resources roughly equivalent to three hours of ChatGPT Pro thinking time.
To shape its disclosure strategy, OpenAI consulted with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study. The company drew upon the advisory group’s advice and public recommendations to establish best practices for the release. While GitHub serves as the initial distribution platform, OpenAI continues to evaluate other community-hosted alternatives that adhere to the committee’s publication guidelines.
Alongside the repository materials, OpenAI announced plans to fund an upcoming series of conferences, workshops, and special academic programs focused on analyzing and understanding major scientific results produced by artificial intelligence systems. OpenAI indicated that it will disclose further details about this funding initiative in the near future as part of its broader stated goal to empower scientific researchers directly with state-of-the-art AI capabilities.
The frontier model that generated these proofs has not yet been released to the public, as OpenAI reported that it is currently working to release the underlying system responsibly. The organization also identified areas for continued development in future research releases, committing to improve the mathematical exposition, citation quality, and overall presentation of papers for better human understanding. OpenAI stated it will continue testing internal frontier models on mathematics and other sciences while adjusting disclosure standards according to feedback from the academic community.
