OpenAI is preparing to launch the new Astra model series for long-term multi-agent collaboration.

Huo Xing Cai Jing reports that on August 1, according to three sources familiar with the matter, OpenAI is preparing to launch a new model series tentatively named "Astra," which will feature enhanced capabilities for completing long-duration tasks. OpenAI’s founder, Altman, presented Astra to policymakers and regulators in Washington this week. OpenAI is said to promote Astra’s ability to enable multiple agents to collaborate over extended periods to solve particularly complex problems, such as large-scale projects or advanced mathematical challenges. Astra will represent a new class of models following OpenAI’s Sol, Terra, and Luna. It remains unclear when OpenAI plans to release the model, and the company has not yet decided whether Astra will be labeled as GPT-6 or as an extension of the GPT-5 series (e.g., GPT-5.7). According to one source, although these models are already in testing, they are expected to be among the first to be submitted under a new framework proposed by the Trump administration, which requires AI models to be disclosed to the federal government prior to public release. The government has set a self-imposed deadline for this weekend to finalize the framework. It is also reported that, in the short term, OpenAI plans to release a report detailing how it solved ten previously unsolved mathematical problems, demonstrating the capabilities of its most advanced AI. (The Information)