On Tuesday (local time), OpenAI made a groundbreaking announcement: its next-generation internal model has triumphantly tackled a Millennium Prize Mathematics Problem associated with the Navier-Stokes equations. This model boasted an impressive pass rate of nearly 48% on a selection of open mathematical problems, nearly tripling the peak performance of GPT-6 Astra. To achieve this feat, the research team harnessed the power of approximately 10,000 collaborative AI agents and dedicated around 88 hours to uncover the solution, with computational expenses soaring into the millions of dollars.

In reaction to this news, Tristan Buckmaster, a mathematician from New York University, raised queries regarding whether OpenAI might have drawn inspiration from his AI-assisted research conducted alongside Levent Alpöge, a researcher at Anthropic, thereby adopting this uncommon methodology. OpenAI promptly responded, clarifying that they had no prior access to the other party's work before its public disclosure and emphasized the substantial disparities in the proof techniques employed by both teams.