Tech
Gist from Wired

OpenAI Launches GPT-6 Astra, Claims We're Already in the AGI Era

Summarized September 4, 2026
Jump to key takeaways

OpenAI has unveiled GPT-6 Astra, its next-generation AI model, billing it as the world's best system for navigating computers, writing code, and solving advanced mathematics. The model is rolling out first to enterprise clients in OpenAI's Daybreak early access program, then to ChatGPT Plus, Pro, Business, and Enterprise subscribers — with no word yet on whether free users will get access.

The launch comes loaded with extraordinary claims from OpenAI leadership. President and co-founder Greg Brockman told reporters he personally believes humanity has already crossed into the AGI era — artificial general intelligence, meaning AI systems that are broadly smarter than humans across tasks. Brockman suggested that when historians look back, GPT-6 Astra's release window will be seen as the inflection point. It's a bold declaration that would have seemed like science fiction even two years ago.

On raw capability, OpenAI says GPT-6 Astra solves a long-standing pain point in AI: reliable computer use. Prior models could browse the web or fill out forms only fitfully. GPT-6 Astra, the company claims, can book DMV appointments, hunt for apartments, and search job listings faster than the average human — tasks that require chaining together many browser interactions without losing context or making errors.

Safety is front and center, but so is a candid admission of growing difficulty. Chief Scientist Jakub Pachocki disclosed that OpenAI is actively monitoring GPT-6 Astra's chain-of-thought reasoning — the internal scratchpad the model uses to work through problems. But Pachocki warned that as models grow more capable, keeping tabs on their internal logic becomes harder. He stated directly that OpenAI would pause further scaling if its ability to monitor model alignment degraded beyond an acceptable threshold — a remarkable public commitment to a potential self-imposed brake on development.

The timing is strategic and pressured. OpenAI is racing toward an IPO while fending off Anthropic, its most formidable rival, which is also preparing to go public. GPT-6 Astra is meant to cement OpenAI's technical lead and supercharge enterprise sales at exactly the moment investor scrutiny is highest.

Key Takeaways

  • Greg Brockman declares the AGI era has already begun
  • GPT-6 Astra: OpenAI's claimed 'world's best computer use model'
  • Model books DMV appointments, job hunts faster than humans
  • Chief Scientist warns monitoring AI alignment is getting harder
  • OpenAI may pause scaling if internal monitoring confidence drops
  • Launch timed ahead of IPO amid fierce rivalry with Anthropic
  • Rollout starts with enterprise Daybreak clients, not free users
Read original article at Wired

Summarize any article in seconds

Gist is a free AI reader for your browser, iPhone, and Android. Get concise summaries and key takeaways from any article or podcast.

Get Gist — Free
⚡ Instant summaries 💬 Chat with articles 🔒 Privacy-first