OpenAI is rolling out GPT-6 Astra, a new model it describes as its most powerful and most “aligned” to date, with strong performance on tasks that involve using a computer. Several outlets report that Astra is focused on agentic, “computer-use” workflows such as operating through interfaces, drafting work, coding, and other professional tasks, and that it is being released gradually to users and developers.

Outlets also emphasize the model’s cybersecurity capabilities. Bloomberg and TechRadar say OpenAI highlights Astra’s ability to handle security-related tasks, while also implementing or tightening guardrails for requests that could enable misuse. TechRadar reports Astra was paused after triggering safety protocols and is now released with restrictions that may be loosened for vetted users. Fortune and Quartz add context by noting broader debate among outside experts about whether OpenAI has sufficient safety and security controls for a model that can do more with less guidance.

Other coverage focuses on user access and rollout details, including how people can get access when it becomes available in ChatGPT plans and via the API and cloud partners. Overall, the reporting aligns on the launch and capabilities, while differing mainly in how they frame the risks, accessibility, and confidence in OpenAI’s safety approach.