OpenAI Claims Next-Gen AI Solved 10-Year-Old Math Problems, Teases 'Astra' Model

OpenAI has unveiled a significant breakthrough, announcing that its upcoming artificial intelligence model, Astra, has successfully solved 10 complex mathematical problems that have stumped researchers for at least a decade. The company emphasized that this achievement demonstrates AI's evolving capability, moving beyond mere research assistance to actively generating novel mathematical insights. The problems tackled spanned advanced fields such as high-dimensional geometry, coding theory, and quantum complexity. OpenAI detailed that Astra's reasoning process for each problem was documented and formalized using 'Lean,' an interactive theorem prover, allowing for computer verification of the AI's proofs. The total computational cost for solving these 10 problems was reportedly around $2,000, based on the API pricing of their recent 'GPT-4.5 Turbo' model. The exact positioning of Astra, whether as an extension of the GPT-4.5 series or as part of a future GPT-5 generation, remains undecided. Reports indicate that OpenAI CEO Sam Altman has already demonstrated Astra to policymakers and regulators in Washington D.C., potentially positioning it as an early subject for the AI safety verification framework being considered by the U.S. administration.

“

Astra has produced new results on 10 unsolved problems in mathematics, which have not progressed for more than 10 years, and most for decades.