OpenAI is reportedly building Astra, a model family designed to work on problems for hours or days

OpenAI is reportedly developing a model family called Astra for long-running tasks that are coordinated across multiple agents.

Astra is still in testing, with no final launch date or branding decided.

Sam Altman reportedly showed the system to policymakers in Washington, D.C.

OpenAI also plans to release a report on its strongest model solving ten previously unsolved math problems.

If successful, Astra could improve long-horizon reasoning and self-correction, and may be among the first models reviewed under a new U.S. federal framework.