OpenAI says it is releasing a broad range of mathematical results produced by an unnamed internal frontier model. In Sharing AI progress in mathematics, published by OpenAI on October 6, 2026, the company describes a release built around a GitHub repository, computer-checkable proof formalizations and additional information about how the results were produced.

The results are in OpenAI's math repository on GitHub. As of November 11, 2026, its README lists 721 manuscripts organized into 373 families, says the model was posed approximately 4,000 problems and says about 42% of top-line results are formalized in Lean. The README describes the manuscripts as "a form of raw research data, rather than polished academic papers," says the collection includes results at different stages of verification and says some of the unformalized results "could have issues." On October 7, OpenAI withdrew three manuscripts because a sign error in one of them invalidated an argument and a construction that the other two depended on, according to the repository's history.

What OpenAI says it is releasing

OpenAI describes the material as a broad set of new mathematical results generated by an internal frontier model. The announcement presents this as a research release rather than a public launch of the model itself. OpenAI says it is working toward responsibly releasing the model that produced the results, and the repository's README calls it an unreleased internal OpenAI model.

The page also says OpenAI consulted the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study while developing practices for sharing the work. That consultation is presented as part of the release process, not as independent confirmation of the mathematical results.

How the results will be shared and checked

For this release, OpenAI says it is publishing the results in a GitHub repository with protocols for revising papers and handling citations. The company says it is also exploring other community-hosted options that meet the advisory group’s guidelines.

The repository includes formalizations of many of the proofs in Lean. Lean is a programming language that allows mathematical proofs to be represented in a form a computer can check. A Lean formalization can make the logical steps of a proof more explicit and testable. The repository's Lean library and formalization catalogue list the available formal proofs and the papers they belong to.

OpenAI says it will add more formalizations as they become available. It also says future releases should improve the papers’ citations, mathematical exposition and presentation so readers can understand the results more easily.

What transparency information accompanies the papers

Alongside the mathematical material, OpenAI says the repository includes additional details about how the results were obtained. These include:

  • 10 summaries of the model’s reasoning

  • Estimates of compute expressed in terms of ChatGPT Pro usage

  • Statistics on the number of problems the model attempted

OpenAI says the average result used compute equivalent to roughly three hours of ChatGPT Pro thinking. That figure is an estimate expressed through a usage comparison, not a detailed account of the hardware, runtime or computational budget for each result.

The repository calls the reasoning summaries "abridged summaries of the model's reasoning." They cover ten results, including the irrationality exponent of π, the symmetric and general Mahler conjectures and Kaplansky's direct-finiteness conjecture in characteristic two.

What comes next

OpenAI says it wants the release to support scientific openness and further progress in mathematics. It plans to fund workshops, conferences and special programs focused on understanding major results produced by AI, and says it will use community feedback to update its standards for future disclosures.

The value of the work will depend on the specific results, the quality of their exposition, the available formalizations and how mathematicians evaluate them.

Sources