GPT-6 Astra Is Here—and OpenAI Thinks It May Kick Off the AGI Era

Maxwell Zeff reports that OpenAI announced GPT-6 Astra on Thursday as its next-generation model, claiming state-of-the-art computer and browser navigation, software writing, and hard math. Paid access starts with Daybreak early-access organizations, then ChatGPT Plus, Pro, Business, and Enterprise; free ChatGPT availability was not confirmed.

In the press briefing, president and cofounder Greg Brockman said it is not unreasonable to feel the world is now in the AGI era, and that looking back in a couple of years, people may date AGI's creation to around this time and this model. The company calls Astra the world's best computer-use model and says it booked DMV appointments, searched job listings, and apartment-hunted faster than the average person in tests. Chief scientist Jakub Pachocki said OpenAI monitors the model's chain of thought for safe deployment, and warned that as capabilities rise, monitorability gets harder — and that confidence in monitoring may constrain further scaling.

The product claim and the AGI rhetoric are doing different jobs. Computer use that actually books appointments and fills forms is the commercial wedge: the industry has long promised agents that replace browser grind and mostly shipped demos. If Astra holds up outside OpenAI's tests, Anthropic and everyone else selling coding-plus-agent stacks have a new bar to clear before their own IPO clocks.

Signal Desk's read: take the computer-use and coding pitch seriously as a competitive move; treat Brockman's AGI line as intentional narrative, not a measurement. "I think we might be about this model" is a founder speaking into a capital and talent market, not a benchmark table. The more consequential sentence in Zeff's piece is Pachocki's: OpenAI says it would withhold further scaling if alignment monitoring confidence degraded past a threshold. That is either a real brake or a useful public commitment. Watch which one it is when the next capability jump arrives.

The timing is the other half of the story. OpenAI needs revenue and a technology story ahead of a planned IPO while competing with Anthropic on the same track. Shipping "most safe and aligned yet" beside "AGI era" is an attempt to hold both narratives at once. Readers should separate the demoable agent work from the spiritual framing. One can be evaluated next week in Daybreak and paid ChatGPT. The other cannot.

OpenAI is pushing capability and sales ahead of a planned IPO while Anthropic prepares for one too. Free-tier access for Astra was left unspecified in the Wired account, which itself is a quiet product decision worth noticing.