OpenAI's new GPT-6 Astra model claims to outperform humans at computer and web tasks. The rollout targets paid users first, as the company pushes automation and faces new safety concerns.
OpenAI has introduced GPT-6 Astra, its most advanced AI model to date. The company says Astra can handle tasks that once required human skill-navigating computers, booking appointments, and searching for jobs online. OpenAI is not pitching this as a minor upgrade, but as a step toward artificial intelligence that can match or surpass people at the core routines of digital work.
OpenAI officially released GPT-6 Astra on September 3, 2026, calling it the most powerful model ever widely deployed and the first to reach Critical-level cyber capabilities under its Preparedness Framework.
The launch comes at a tense moment for OpenAI. The company is moving quickly to develop new products and boost revenue ahead of a planned IPO, while its main rival, Anthropic, is preparing to go public as well. OpenAI’s Daybreak early access program will be the first to try GPT-6 Astra, followed by ChatGPT Plus, Pro, Business, and Enterprise subscribers. Free users have not been given a timeline for access.
OpenAI claims GPT-6 Astra is the "world’s best computer use model." In controlled tests, the AI booked DMV appointments, searched job listings, and found apartments faster than the average person. For publishers and digital creators, this could mean a new wave of automation: tasks that once needed manual work or custom scripts may soon be handled by a model that can use software tools quickly and accurately. According to a Reuters report on the Astra launch, the model demonstrated skills like apartment hunting, tax preparation, and legal memo formatting, showing its range across professional fields.
OpenAI has stated that Astra is initially available only to a limited group of organizations, with broader access for ChatGPT Plus, Pro, Business, and Enterprise subscribers, as well as via API and AWS, rolling out in the following days. The free tier was not listed as receiving immediate access in the official announcement.
For digital publishers, Astra’s ability to automate browser navigation and software use could speed up editorial work, research, and how content is gathered and distributed. But these same features raise concerns about security and oversight. As OpenAI's own safety overview notes, Astra can spot unknown security flaws and, if given the right tools and access, find ways to exploit them in protected systems. Because of this, advanced cybersecurity features are limited to testers for now and will expand through the Daybreak Blue program.
OpenAI is restricting early access to paying customers and enterprise partners, signaling that the most powerful AI tools will go first to those willing to pay. This approach brings in revenue and lets OpenAI monitor how Astra is used before opening it up to the wider public. The message to competitors and creators is clear: those with resources and oversight will shape the next phase of digital work.
GPT-6 Astra’s ability to automate complex digital tasks could boost productivity for publishers and creators, but it also puts more power in the hands of those with early access. OpenAI’s public optimism about entering the AGI era is balanced by private caution about safety and control. For now, enterprise clients and paid subscribers will be the first to use the new model. Free users will have to wait as the bar for entry-and the risks of unchecked automation-rise. OpenAI’s next steps will show whether this leap forward strengthens its lead or exposes new weaknesses in the race to automate digital work.
OpenAI was founded in 2015 as a nonprofit research lab. By 2026, it employs over 1,500 people and reports annual revenues above $2.5 billion, mostly from its GPT models and enterprise AI services. Its main product, ChatGPT, passed 200 million registered users earlier this year, making OpenAI a major player in commercial AI.
