OpenAI Launches GPT-6 Astra, Claims It May Mark the Dawn of the AGI Era

OpenAI has unveiled GPT-6 Astra, its next-generation AI model, which the company is billing as the world's best model at computer use — capable of navigating browsers, booking DMV appointments, hunting for apartments, and searching job listings faster than the average human. The rollout begins with a limited set of enterprise clients in OpenAI's Daybreak early access program before expanding to ChatGPT Plus, Pro, Business, and Enterprise subscribers. Free users remain in limbo, with no announcement on their access.

The launch carries extraordinary weight internally. OpenAI president and cofounder Greg Brockman told reporters he personally believes the world has already crossed into the artificial general intelligence era — systems that are, in a general sense, smarter than humans. He suggested that when historians look back, GPT-6 Astra may be identified as the inflection point. It's a bold claim, and one that puts OpenAI's credibility squarely on the line.

The timing is strategically charged. OpenAI is racing to demonstrate technological leadership and accelerate revenue ahead of a planned IPO, while fending off Anthropic — its closest rival, which is also pursuing a public offering. GPT-6 Astra is meant to signal dominance on both fronts: cutting-edge capability and commercial viability through its computer-use features that could automate a wide range of white-collar tasks.

But even as OpenAI celebrates the model's capabilities, its own chief scientist flagged a growing and potentially serious constraint: as models become more powerful, monitoring their internal reasoning becomes harder. Jakub Pachocki warned that the company's ability to observe GPT-6 Astra's chain-of-thought — its internal problem-solving scratchpad — is already being tested. He stated plainly that if OpenAI loses sufficient confidence in its ability to monitor model alignment, it would halt further scaling rather than proceed blind. That's a notable admission: the safety challenge may soon throttle the very progress OpenAI is celebrating.