OpenAI began rolling out GPT-6 Astra on Sept. 3, calling it "the world's most intelligent and aligned model" in its announcement, according to 9to5Mac's report on the launch.
OpenAI said the model sets a new standard for computer use, browsing, software engineering, cybersecurity and scientific work, and that ChatGPT is nearly twice as fast at computer use tasks as before. For Codex, Astra can keep notes across context windows instead of compressing older information away, preserving accumulated details across long coding sessions, the company said.
The model ships with a 1.05 million token context window and a maximum output of 128,000 tokens, with a knowledge cutoff of April 30, 2026. OpenAI is pricing API access at $10 per million input tokens and $50 per million output tokens.
Initial access went first to enterprise customers in OpenAI's Trusted Access Program and to companies in its application-based cybersecurity program, 9to5Mac reported. OpenAI said the model would then expand to ChatGPT Plus, Pro, Business and Enterprise plans and to the API and Amazon Web Services over the following days, with enterprise administrators needing to turn access on since it is off by default.
A model that can operate a browser and a terminal on its own is a bigger deal for how software gets built than another jump in a coding benchmark. Teams already running agents against OpenAI's API should expect the gap between what Astra can do unsupervised and what their current guardrails assume to be wider than it was a week ago.