OpenAI named its next major model on Saturday, and it did so in the third paragraph of a research post about maths. Astra is built to stay on a single problem for hours or days by splitting the work across several agents that talk to each other. An internal version of it produced ten results in mathematics and theoretical computer science, each one shipped with a proof a machine can check. The release date, the price and the final name all stay open.What OpenAI actually saidThe company published a post titled "Ten advances in mathematics and theoretical computer science" and buried the reveal inside it. The results came from an internal version of Astra, described as "our next major model," and the ten problems selected had sat still on their main result for at least a decade, in most cases far longer. The token cost of finding the solutions would run to roughly $2,000 at Sol API rates. Humans then prepared the arguments into manuscripts alongside the same model, after which the model wrote each argument up as a Lean certificate, and OpenAI released the model's own narration of how it got there.About The AuthorAt heart, I am a storyteller drawn to the watershed moments that bend the technology landscape. I braid narrative with data, humanise statistics, and trace the arc from first spark to world-changing impact. My reportage, features and reviews are witty, sardonic, visual and vivid, using anecdote to illuminate rather than eviscerate.
OpenAI Astra Runs For Days And Cracked 10 Maths Problems
Astra is OpenAI's next major model family, announced on 1 August 2026, built to run several agents on one problem for hours or days. An internal version produced ten new results in mathematics and theoretical computer science, each shipped with a machine-checkable Lean 4 proof. The release date, price and final name stay open.
OpenAI unveiled Astra, solving 10 decade-old mathematics problems via multi-agent coordination, verified with formal Lean proofs. For enterprise leaders, this signals OpenAI's move from single-turn chat to sustained reasoning—critical for complex problem-solving and AI strategy.










