OpenAI's GPT-6 Astra Claims Human-Level Computer Control

OpenAI releases model designed to operate computers autonomously

OpenAI launched GPT-6 Astra on Thursday, positioning the model as a breakthrough in AI's ability to control computers and web browsers on behalf of users. The company claims the system can book DMV appointments, search job listings, and hunt for apartments faster than the average person performing the same tasks.

The rollout begins with enterprise customers in OpenAI's Daybreak early access program, followed by ChatGPT Plus, Pro, and Business subscribers in the coming days. The company has not announced whether free ChatGPT users will gain access to the new model.

Why it matters

Reliable computer control represents a longstanding challenge in AI development. If GPT-6 Astra performs as advertised, it could automate routine digital tasks that have resisted previous automation attempts—from filling out forms to navigating complex software interfaces. The technology arrives as OpenAI prepares for a planned initial public offering and faces intensifying competition from Anthropic, which is also pursuing an IPO.

Executives signal confidence in AGI timeline

During a press briefing, OpenAI president and cofounder Greg Brockman suggested the industry has entered the era of artificial general intelligence—systems that match or exceed human cognitive abilities across domains.

"It's not unreasonable to feel that we are now in the AGI era," Brockman told reporters. "I think that if we fast forward a couple of years, when we look back and say, 'When was it really that AGI was created?' I think it's going to be about this time, and I think it might be about this model."

OpenAI describes GPT-6 Astra in a blog post as the "world's best computer use model," claiming substantial improvements over earlier versions in navigating websites, completing forms, and manipulating software tools.

Safety monitoring emerges as potential constraint

Chief scientist Jakub Pachocki told reporters that OpenAI monitors the model's chain-of-thought—its internal reasoning process—to ensure safe deployment. However, he acknowledged that maintaining visibility into increasingly capable systems presents growing difficulties.

"We kind of take this visibility for granted, and we are seeing that as model capabilities are increasing, monitorability is getting more challenging," Pachocki said. He indicated that the company's ability to monitor model alignment may limit future scaling decisions.

"We think confidence in monitoring may constrain further development, because we would not accept degradation in our ability to monitor model alignment beyond a certain level," Pachocki explained. "We would withhold scaling until we can regain enough confidence."

Despite these concerns, OpenAI maintains that GPT-6 Astra represents its most safe and aligned model to date.

The details were first reported by WIRED.