OpenAI unveils GPT-6 Astra, claiming arrival of the AGI era
The new model targets enterprise workflows and coding, with a rollout timed to precede a planned initial public offering.
OpenAI has announced the launch of GPT-6 Astra, its next-generation artificial intelligence model, which the company describes as state-of-the-art in computer navigation, software coding, and complex mathematics. In a press briefing, President and co-founder Greg Brockman stated that he personally believes the world is now entering the era of artificial general intelligence, or AGI. Brockman suggested that looking back in a few years, this period and this specific model would likely be identified as the moment AGI was created.
The rollout is set to begin in the coming days, initially targeting a limited set of organisations, including enterprise clients in the Daybreak early access program. Availability will subsequently expand to subscribers of ChatGPT Plus, Pro, Business, and Enterprise plans. OpenAI has not yet confirmed whether the model will be made available to free users of the platform.
The timing of the release is significant for investors, as OpenAI seeks to rapidly advance its technology and boost sales ahead of a planned initial public offering. The company is also facing stiff competition from Anthropic, its most important rival, which is also preparing for an upcoming listing. OpenAI claims GPT-6 Astra is the “world’s best computer use model,” a capability that AI companies have historically struggled to perfect.
In internal tests, OpenAI reported that GPT-6 Astra was able to book DMV appointments, search for job listings, and hunt for apartments faster than the average person. These results address a long-standing industry challenge: creating models that can reliably browse websites, fill out forms, and manipulate software tools with human-like speed and accuracy.
Despite the performance claims, questions remain regarding the model’s broader reliability. The assertion that GPT-6 Astra marks the start of the AGI era is based on OpenAI’s internal assessment and Brockman’s personal belief, rather than an independent scientific consensus. It remains unclear if the model’s success in specific tasks will translate to consistent reliability across all user scenarios.
OpenAI has labelled GPT-6 Astra its safest and most aligned model to date. However, chief scientist Jakub Pachocki noted that monitoring the model’s “chain-of-thought” is becoming a bottleneck to further progress. Pachocki explained that as model capabilities increase, monitorability is getting more challenging, and the company may withhold scaling until it can regain enough confidence in its ability to monitor model alignment.
