OpenAI has revealed Astra, an unreleased model designed to tackle complex, long-running tasks, after an internal version produced ten significant advances in mathematics and theoretical computer science.
In a new research post, OpenAI described Astra as “our next major model” and said the problems had seen no progress on their central results for at least a decade, and in most cases, much longer.
According to OpenAI, its internal research focused on a wide range of areas, including high-dimensional geometry, coding theory, arithmetic circuit complexity, group theory, quantum complexity, lattice cryptography, and extremal combinatorics.
OpenAI also says it’s advancing rapidly in science.
Some examples include the existence of non-sofic groups, a disproof of Connes’s rigidity conjecture, new bounds for high-dimensional sphere packing, and results resolving several problems posed by mathematician Paul Erdős.
“The total number of tokens needed to find solutions to these problems would cost roughly $2,000 at Sol API rates,” OpenAI noted.
Human researchers used the same model to prepare the arguments as manuscripts.
Astra then formalized every argument as a Lean certificate, allowing the proofs to be checked using the mathematical verification system.
Astra could launch as GPT-5.7 or GPT-6
The Information also independently confirmed that OpenAI is indeed working on Astra, a new model family built for long-running workloads.
As per OpenAI, Astra is a powerful model that allows AI agents to collaborate on different parts of a larger problem.
BleepingComputer understands that OpenAI has reportedly not decided whether the model will be released as GPT-5.7, GPT-6, or under another name.
At this point, we know that Astra qualifies as a major breakthrough AI model, and it could be subject to Anthropic-like policies where one version is released to consumers, while a more powerful variant requires special approval.
Security teams log 54% of successful attacks and alert on just 14%. The rest move through your environment unseen.
The Picus whitepaper shows how breach and attack simulation tests your SIEM and EDR rules so threats stop slipping by detection.