OpenAI has released a large batch of mathematical results produced by an internal frontier model, claiming progress or solutions on hundreds of open problems across major areas of mathematics while largely avoiding a high-profile media rollout.

The company published hundreds of manuscripts, reports put the total at more than 700 papers organized into roughly 370 result families, on GitHub. The findings span algebra, number theory, theoretical computer science, mathematical logic and topology. Many of the proofs were formalized in the Lean verification language. OpenAI said the average result required the equivalent of about three hours of ChatGPT Pro compute. The release follows the company’s earlier claim of progress on the Navier-Stokes equations, one of the Millennium Prize Problems.

Among the new results are claimed advances related to the Riemann hypothesis, a solution to the four-dimensional Kakeya conjecture, and progress on other long-standing questions. OpenAI provided limited chain-of-thought summaries for a small number of the results and did not publicly name the specific model used.

The scale of the dump has stunned parts of the mathematics community. Researchers note that AI systems are now solving problems that individual mathematicians spent decades pursuing. Some mathematicians have openly described the technology as dethroning humans as the primary theorem-proving force, with one prominent researcher writing that human mathematicians are “forevermore dethroned as the main theorem-proving entities on planet earth.”

OpenAI framed the release as a byproduct of model evaluation after existing math benchmarks became saturated. The company said it is funding workshops and programs to help the community understand the results and is committed to improving paper quality in future releases. Critics argue the low-key GitHub approach and incomplete formalizations for many papers leave mathematicians to do the hard work of verification while the company advances proprietary systems that few outside labs can match.

The episode underscores a deeper shift: frontier AI models are increasingly capable of generating novel mathematical results at a pace and scale that traditional academic research cannot match, raising questions about the future role of human mathematicians even as OpenAI presents the work primarily as scientific progress rather than a direct challenge to the profession