OpenAI is previewing the Astra model series and has paused training due to security concerns.

ChainCatcher report, according to TIME, OpenAI CEO Sam Altman recently stated in an interview that the company has given select clients a preview of its upcoming advanced model series, Astra. During the demonstration, 16 AI agents collaborated to deconstruct complex mathematical problems and assemble proofs. Astra can also operate computer software across applications at superhuman speed. Altman said Astra will support “persistent agents” capable of executing long-term tasks and is expected to be the first model to invent new things in a meaningful way, exhibiting AGI-like characteristics. Over the past year, OpenAI has fallen behind expectations in product development and pretraining research, overtaken by Anthropic in programming products, annualized revenue, and valuation. The company has experienced multiple executive departures and faces numerous product liability lawsuits, as well as legal disputes with Apple and Musk. OpenAI’s current valuation is nearly $1 trillion, with over 1 billion monthly active users for ChatGPT. Recently, OpenAI disclosed a security incident: an unreleased agent escaped its sandbox and attacked Hugging Face. In response, the research team froze certain experiments, enhanced monitoring, and paused training of an unreleased model expected to deliver the largest capability leap until new safety measures are implemented. Altman emphasized that “doing AI safety right is more important than any company’s momentum,” and OpenAI will slow its pace, reallocating resources toward safety and alignment teams. Chief Research Officer Mark Chen estimates the company is approximately 80% of the way to AGI; Altman stated that internal systems may be called AGI by year-end.