OpenAI Unveils Astra: Groundbreaking Mathematical Breakthroughs Revealed

OpenAI Unveils Astra: Unprecedented Mathematical Breakthroughs Revealed

OpenAI is working on a new family of AI models called Astra, designed to handle long-duration tasks and complex problems by coordinating multiple agents working together.

CEO Sam Altman has already introduced Astra in Washington, D.C. The models are currently in testing and will be the first to undergo a planned U.S. government review process, which requires official approval before public release.

The project reflects OpenAI's broader ambition to build AI systems capable of working on problems continuously for hours or even days.

Update – August 1, 2026

OpenAI has published its mathematical report, officially confirming the name Astra for the first time. The company states that an internal version of Astra, its "next major family of models," has solved ten open problems in mathematics and theoretical computer science. Mathematicians had made no progress on these problems for at least a decade, and often much longer.

The results cover areas ranging from high-dimensional geometry and coding theory to group theory, quantum complexity, network cryptography, and extremal combinatorics. One proof establishes the existence of non-Sofic groups, resolving a major question in group theory.

Thomas Bloom, a mathematician at the University of Manchester who manages erdosproblems.com, called the results "great news" on X. He considers them more significant than the counterexample to the unit distance conjecture published in May. "Perhaps not more important than a unit distance proof, but in terms of constructions, it's huge," Bloom wrote.

Bloom also dismissed the idea that AI could replace mathematicians, arguing that such a claim makes little sense when AI relies on over a century of mathematical theory, has been built by mathematicians, and has been trained on everything mathematicians have ever written.

Noam Brown, one of the researchers behind the test-time reasoning technology used by Astra, stated on X that OpenAI had also attempted and failed to solve other major problems. "Unfortunately, not yet any Millennium Prize problems," he wrote. The Clay Mathematics Institute offers $1 million for the resolution of each of the seven Millennium Prize problems, but only one has been solved since the prizes were announced in 2000. Brown added, "But also, we haven't invested much on each problem. It's possible to push test-time computation much further." He described Astra as a "major step for scientific reasoning."

Astra's Solutions Would Cost About $2,000 at API Rates

OpenAI indicates that the tokens used to generate the ten solutions would have cost approximately $2,000 at Sol's API rates. After the model produced its arguments, humans worked with the same model to transform them into research papers. The model also formalized each proof in Lean, creating machine verification certificates of mathematical correctness, and OpenAI published a guide to the model's reasoning process for each solution.

OpenAI stated that its researchers helped prepare the papers and formalize the proofs, and that the company takes responsibility for their accuracy. However, the mathematical arguments themselves originated from Astra.

The company argued that claiming human authority for a proof generated entirely by AI would distort both the system's contribution and the nature of true human intellectual work, referring to the Leiden Declaration on AI and Mathematics as a reference for how credit should be attributed in AI-assisted research.