OpenAI releases internal model results

OpenAI uploaded 719 manuscripts on October 6, 2026, created by an internal model that was never used commercially. The manuscripts are grouped into 372 families of results and claim to solve hundreds of open problems, some of which had gone unsolved for decades.

Mathematician reacts to the release

French mathematician Hugo Duminil-Copin, a Fields Medalist in 2022, described his field of research as "annihilated" after learning that some of his favorite questions were among the results. He shared with colleagues the emotional impact of the discovery and the challenge of reviewing thousands of pages of material.

Computational effort behind the results

OpenAI stated that it submitted around 4,000 problems to the model, with each result requiring the equivalent of three hours of computation by ChatGPT Pro. The company emphasized the significant computational resources involved in generating the manuscripts.

Verification of the proofs

According to Le Monde, 42% of the proofs in the manuscripts were verified using Lean, a software that checks proofs line by line. However, the remaining proofs may contain errors, and at least three articles have already been retracted, as reported by Fortune.

Duminil-Copin’s response to the challenge

Duminil-Copin estimates he has more than 3,500 pages to review on his areas of study. He admits he does not want to read the proofs, as they lack the personal touch of texts written by colleagues he knows. Nevertheless, he plans to engage with the material to help the mathematical community understand and transmit the findings.

OpenAI’s support for the community

OpenAI has pledged to fund workshops and conferences to help mathematicians assimilate the results. It is also consulting an independent group at the Institute for Advanced Study in Princeton to guide the process.

Expert reaction to the impact

Cédric Villani, a Fields Medalist in 2010, called the release "the most radical change in mathematics since at least four centuries." He estimates that the results could yield a dozen new Fields Medals, highlighting the potential significance of the work.

Previous claims by OpenAI

OpenAI had previously claimed a solution to the Navier-Stokes problem, one of the seven "Millennium Problems" each worth one million dollars, at the beginning of September. This problem is considered one of the most important unsolved questions in mathematics.