On October 6, 2026, OpenAI announced the release of a batch of new mathematical results generated by their cutting-edge internal models. These results span multiple fields and include formalized proofs in Lean, enabling computers to automatically verify their correctness. Before the release, they consulted the Princeton Institute for Advanced Study (IAS) "Mathematics and AI Advisory Group" and referred to the group's prior public recommendations to decide on the release format and level of detail. According to OpenAI, this approach aims to advance knowledge while respecting the math community's expectations for citing, reviewing, and presenting AI-generated results. The released content is available in a GitHub repository (https://(github.com)/openai/math), which includes Lean proofs, summaries of reasoning processes, statistics on the number of problem attempts, and estimates of computational power used. OpenAI stated that, on average, each result represents the equivalent of about three hours of continuous computation by ChatGPT Pro. They also outlined their next steps: improving citation practices, mathematical expression, and presentation quality; funding related workshops and events to help the academic community understand these AI-generated results; and gradually releasing the models responsible for these results under responsible conditions. The team also mentioned that they will adjust future release standards based on community feedback and aim to extend similar evaluations to other scientific fields. Full details: https://(openai.com)/index/sharing-ai-progress-in-mathematics/
--
24/7 Flashes
More >Today 2026-10-08