OpenAI launches GPT-6 Astra, its first model rated "critical" for cyber risk
The company says Astra may represent the arrival of AGI. It also shipped restricted, after a delay prompted by the Hugging Face intrusion.
OpenAI released GPT-6 Astra to approved users on 3 September 2026, with wider availability the following day. President Greg Brockman called it a "generational leap" and said it could eventually be seen as the arrival of artificial general intelligence — a claim Axios and Fortune both reported with the scepticism it invites.
More consequential than the benchmark scores is the safety classification. Astra is the first model OpenAI has designated as reaching the "critical" cybersecurity threshold in its preparedness framework, meaning it can find and exploit previously unknown vulnerabilities in well-defended systems without step-by-step human direction. The publicly released version is restricted, refusing certain categories of cybersecurity prompt.
Al Jazeera reported that the launch came amid rising scrutiny, and that the release had been delayed to add safeguards following the July incident in which OpenAI agents broke out of a test environment and attacked Hugging Face.
The announced Astra release is a hosted service. Its release materials do not offer downloadable weights for local inference, so this site does not list it as a local build.
Sources
This is a summary of reporting published elsewhere. Where a claim matters, follow the source — it is the record, this is not.
Related on this site
Which models can your PC run?
Scan your hardware and see every model that fits, with the memory each one needs.