OpenAI unveiled GPT-6 Astra exactly one year after it launched GPT-5. Yesterday, the company began rolling GPT-6 Astra out to a small group of organizations.
OpenAI says GPT-6 Astra is its smartest model yet, and the one that’s most closely aligned with user intent. The company says the biggest jump shows up on the hardest professional work: programming, science, and autonomous computer use.
A model so capable that people are already talking about AGI
By OpenAI’s own numbers, Astra scores 98% on FrontierMath Tier 4, 99.9% on ARC-AGI-3, and 100% on ExploitBench. The company is pitching it as a major step up from GPT-5.6 Sol, Terra, and Luna, which were released just two months ago.
OpenAI also says ChatGPT with Astra can use applications, browse the web, and handle multi-step workflows like creating documents, spreadsheets, and presentations. On most tasks, the company says, it’s nearly twice as fast.
Greg Brockman serves as OpenAI’s president, and he believes we’re starting to move into the era of artificial general intelligence. Part of that claim comes from OpenAI’s own testing: Astra’s cybersecurity capabilities have reached the company’s “Critical” level. OpenAI has also put new security measures in place after the Hugging Face incident.
The rollout started yesterday with selected organizations and, OpenAI says, GPT-6 Astra will reach the rest of ChatGPT Plus, Pro, Business, and Enterprise users in the next few days, along with the API and AWS.
