OpenAI says its internal model produced 372 math results, nearly all from a single prompt handed to a single AI agent; some might have taken multiple attempts (Joseph Howlett/Scientific American)

OpenAI has released hundreds more math results from its internal model, following its recent breakthrough in resolving one of the six biggest open problems in mathematics. The company shared the results in a GitHub repository at 6 PM EDT. Many of these results have already been verified in Lean, a programming language that validates proof logic, and are therefore considered likely to be correct.

The new model generated almost all of these results in response to a single prompt given to a single AI agent, which is a stark contrast to the previous Navier-Stokes solution that required a 10,000-strong agentic swarm and significant computing power. While some mathematicians are skeptical of OpenAI's claims due to its history of bold statements and lack of transparency, others believe that making these results accessible to the community will advance mathematical knowledge.

Despite the skepticism, OpenAI insists that it cannot slow down its pace of releasing these results, as they serve as a crucial test to demonstrate their AI's increasing capabilities.