Artificial Intelligence

OpenAI unveils Astra model, claims milestone in artificial general intelligence

OpenAI has launched its new GPT-6 Astra model, with company president Greg Brockman declaring the arrival of the "AGI era", though rival tech leaders and even OpenAI's own chief executive have offered differing views on the term's meaning.

By Jack Douglas | 4 September 2026
3D rendered abstract design featuring a digital brain visual with vibrant colors.

OpenAI has announced the release of its latest artificial intelligence model, Astra, with the company's president declaring that the technology represents the arrival of "the AGI era" — a reference to artificial general intelligence, a long-sought and loosely defined goal within the industry.

Greg Brockman made the comment at the close of a press call on Thursday unveiling the model, which is being rolled out to customers this week. OpenAI defines AGI as "highly autonomous systems that outperform humans at most economically valuable work". Brockman told journalists he believed people would in future look back and identify this period, and potentially this specific model, as the point at which AGI was achieved.

"For me personally, I do think we're there," he said.

The new model, referred to by OpenAI as GPT-6 Astra, arrives roughly a year after the company launched its GPT-5 family of models and around two months after its most recent update, GPT-5.6. OpenAI described Astra as "the world's most intelligent and aligned model" and said it marked "a step change in professional work".

Among the capabilities highlighted by the company is an enhancement to ChatGPT's "computer use" functions, which allow the AI to carry out tasks autonomously on a user's desktop or within an internet browser. According to OpenAI, Astra can fill out forms, format documents, schedule appointments and operate software on a user's behalf.

OpenAI employees told Business Insider last month that these computer-use tools had reached what they described as an inflection point in their development. The ambition to build reliable AI agents capable of navigating everyday computer tasks has been a long-standing target for OpenAI and other major technology firms, though the work has proved difficult given the enormous variation in how people use their computers. Existing tools of this kind have also been criticised for being comparatively slow.

In a post on X, OpenAI wrote: "Anything you can do on a computer, Astra can do for you. Fast."

The company had previously said it delayed Astra's release until September after identifying areas for improvement in the model's cybersecurity capabilities. Earlier this week, OpenAI said it needed to be confident that safeguards built into the system would prevent misuse by malicious actors seeking to exploit it for hacking purposes, and that the model itself was properly aligned to carry out only what users intended.

The term AGI remains without a single agreed definition across the technology sector, and views on when — or whether — it has been reached vary significantly among industry leaders. Demis Hassabis, co-founder of Google DeepMind, said at his company's conference in May that he expected AGI to arrive as soon as 2030, describing humanity as standing "on the foothills of the singularity".

By contrast, OpenAI's own chief executive, Sam Altman, has previously played down the significance of the term. Speaking on the Sources podcast, Altman described AGI as, at best, "a very poorly defined term", adding: "I was going to say it's like an irrelevant marketing term."

The contrast between Brockman's declaration and Altman's more sceptical characterisation of AGI reflects a broader lack of consensus within the industry about what would actually constitute the milestone, and how it should be measured or verified. OpenAI's announcement did not set out specific benchmarks or independent evaluations to support the claim that Astra represents a step into an "AGI era", and the company's characterisation of the model rests primarily on its own internal assessment.

Astra's rollout follows a pattern of rapid iteration by OpenAI, which has released successive model updates at a steady pace over recent years as competition among AI developers has intensified. The company has framed each release as a significant advance in capability, while continuing to face scrutiny over safety, misuse risks and the pace at which such systems are being deployed to the public.

OpenAI has not detailed further technical specifications of Astra beyond the capabilities described in its announcement, and the company said the model would continue to be made available to customers through the course of this week.