OpenAI has formally launched GPT-6 Astra, its newest synthetic intelligence mannequin, describing it as “the world’s most clever” mannequin. The corporate says Astra is designed to deal with complicated duties throughout software program engineering, science, cybersecurity, monetary modelling and information evaluation, whereas additionally taking over sensible work akin to tax preparation and kind filling.
OpenAI President Greg Brockman described the mannequin as a “generational leap in functionality” whereas talking to reporters at a press briefing.
GPT-6 Astra is state-of-the-art on FrontierMath Tier 4, ARC-AGI 3, and TerminalBench-4.0.
GPT‑6 Astra can be a significant advance for scientific discovery, with state-of-the-art efficiency on Terminal-Bench Science 0.1 and HealthBench Professional. pic.twitter.com/7hEFadVAN9
— OpenAI (@OpenAI) September 3, 2026
Brockman additionally addressed the long-running debate round Synthetic Normal Intelligence, or AGI. In response to The Verge, Brockman stated that individuals trying again in just a few years may doubtlessly establish Astra as the purpose when AGI was created.
He went additional, saying, “For me personally, I do suppose we’re there,” earlier than concluding, “Welcome to the AGI period.” OpenAI’s official Astra documentation, in the meantime, focuses on the mannequin’s capabilities, security measures and cybersecurity dangers moderately than formally declaring AGI.
GPT-6 Astra is right here.
We hope it’s going to start to allow a brand new technology of entrepreneurship, scientific discovery, and constructing.
We consider it’s the greatest mannequin on this planet for pc use, skilled work, science, coding, cybersecurity, and extra.
It took us some additional time…
— Sam Altman (@sama) September 3, 2026
Cybersecurity And A Managed Rollout
Cybersecurity is without doubt one of the largest areas of focus for Astra. OpenAI says the mannequin has reached its Vital cybersecurity functionality threshold, which means it might establish and develop useful exploits for vulnerabilities in hardened programs.
The corporate says Astra found two beforehand unknown vulnerabilities throughout inside evaluations, which have been disclosed to maintainers. As a result of these capabilities can be utilized for each defence and exploitation, OpenAI is initially limiting entry to its most superior cybersecurity capabilities, with trusted defenders getting entry by way of the Dawn Blue programme.
In a weblog submit, OpenAI stated Astra can autonomously deal with a variety of “tedious” pc duties, together with web site creation, scientific evaluation, recreation growth, cybersecurity and coding.
For example the time financial savings of constructing autonomous AI brokers with Astra, the corporate stated the mannequin may minimize house searching from six hours to beneath 10 minutes.
Astra Launch Day Additionally Noticed An AI Outage
However on the identical day OpenAI was making ready to roll out Astra, ChatGPT, Claude and Grok all skilled outages. In response to The Verge, ChatGPT customers reported issues with conversations, logins, file uploads, voice mode, search, Deep Analysis and picture technology, whereas Claude and Grok additionally confronted service disruptions.
The precise causes of the simultaneous outages have been unclear, and there was no confirmed proof that Astra triggered the ChatGPT outage. OpenAI says Astra can be made obtainable to ChatGPT Plus, Professional, Enterprise and Enterprise customers, in addition to API builders, within the coming days.
