OpenAI Just Dropped GPT-6 Astra - And Called It the Start of the AGI Era
Astra is OpenAI's most powerful model yet: computer use, coding, cyber. First model to trip internal safety thresholds. Rolling out to Plus, Pro, Enterprise this week.
OpenAI stopped whispering about AGI and just said the quiet part out loud.
GPT-6 Astra launched on September 3 - the company’s most capable model yet, and the first one OpenAI says triggered advanced internal safety protections because of its cyber skills. President Greg Brockman ended a press briefing with four words that will live in quote tweets forever: “Welcome to the AGI era.”
What Astra actually does
OpenAI pitches state-of-the-art computer use, browser use, software engineering, science, and professional work. Demo energy: voice commands that go from a yellow circle to a full 3D game in minutes, then spin up an eBay listing without clicking like a human. Coding scores allegedly beat prior models including GPT-5.6 Sol. Cyber benchmarks are the part that made lawyers sweat.
The company says Astra can find previously unknown security flaws and build exploits across protected systems with less hand-holding than older models. That is why Daybreak - OpenAI’s defender program - gets it first. Paid ChatGPT plans and the API follow over the coming days.
Why people are freaked out and excited at the same time
Because “the model is so good we had to turn on extra locks” is either marketing genius or a genuine milestone. Opaque recurrence - the controversial reasoning method critics keep circling - is part of the story. Altman said the White House was briefed. The internet heard “AGI” and lost its mind on schedule.
The magazine read
New frontier model. Computer use. Cyber thresholds. “Welcome to the AGI era.” Available this week if you pay. Whether Astra is real AGI or just the best chatbot yet, September 3 just became a date people will argue about for years.