OpenAI unveils Astra, first model slated for federal approval before public release

OpenAI presented Astra to U.S. government officials in Washington this week, according to a report by The Information citing people familiar with the company. The model was designed from the ground up to coordinate multiple agents working together over extended periods on exceptionally complex tasks. The company has not yet decided whether Astra will launch a GPT-6 series or slot in as GPT-5.7 within the existing line, and no release schedule has been set.

First federal approval

Astra is expected to become the first model to pass through the new federal approval system established by the Trump administration, which mandates government review before a large language model can be released to the public. In parallel, OpenAI plans to publish a report detailing how it solved ten previously unsolved mathematical problems, an achievement likely enabled by Astra's capabilities.

Celestial theme and regulatory shift

The company's recent model names — Sol, Terra and Luna, which replaced the GPT-5.6 iterations — signaled a celestial naming convention that Astra continues naturally. The new approval framework marks a fundamental regulatory change: for the first time, a large language model must undergo formal federal evaluation before release, rather than relying solely on internal or voluntary review. The forthcoming report on the ten mathematical problems will provide the first public evidence of mathematical-reasoning capabilities that have not been seen in commercial models to date.