OpenAI Publishes 300 Math Solutions Generated by Unreleased Frontier Model

OpenAI published a repository containing solutions to more than 300 mathematical problems on Oct. 6, generated by its internal, unreleased frontier artificial intelligence model. The company announced revised processes for citations and paper revisions alongside the release, acknowledging expectations that experts will scrutinize the results.

"For future releases, we are committed to further improving the quality of the papers via the citations, mathematical exposition, and presentation of the results for better understanding," OpenAI stated in documentation accompanying the repository.

As reported by Mashable Tech , the latest release follows controversy over OpenAI's announcement last month that 10,000 AI agents collaborated to solve the Navier–Stokes existence and smoothness problem—a narrower subset of the unsolved equation within the Millennium Prize Problems set. OpenAI declined a potential $1 million prize for that work. Academics subsequently criticized the approach for addressing only a subset of the problem through what some called an "unnatural" mathematical loophole.

Following that earlier announcement, the Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study stated it did not endorse AI developers using proprietary models to test solutions without scientific community involvement. OpenAI responded by announcing a new advisory group to guide future releases, though it appears to have used the same unreleased model for the new batch of equations.

Unlike the 10,000-agent setup deployed for Navier–Stokes, OpenAI told Scientific American that the latest 300-plus problems were solved using a single prompt given to a single AI agent. Outside mathematicians are already questioning the claim while awaiting access to the proprietary system to verify the work.