Altman says unreleased OpenAI model proved a Millennium Prize Problem

OpenAI CEO Sam Altman said at the ADIPEC opening ceremony in Abu Dhabi that OpenAI used the next GPT model, which is to be released soon, to prove a Millennium Prize Problem.

Rohan Paul, relaying the remarks on X, says 10,000 copies of the model ran nonstop for about four days and worked without hierarchy, each choosing its own direction and sharing progress. The full video is on the ADIPEC Official YouTube channel. The reports do not name the problem or publish a proof, and no independent verification is reported.