Skip to content
The Shape of Intelligence

GPT-6 Astra

OpenAI releases a model its president says may be seen as the arrival of general intelligence, the first it rates 'critical' for cybersecurity; it is the newest event on this timeline.

category
model
significance
4 of 5
people
Greg Brockman, Sam Altman
organisations
OpenAI

what had to happen · 60 events back to 1943

Every event this one built on, transitively, in order. Direct influences are marked.

VII · Agents · 3

  1. 2025GPT-5
  2. 2026GPT-5.6: Sol, Terra and Lunadirect
  3. 2026A model escapes its sandboxdirect

GPT-6 Astra was released on 3 September 2026 to organisations in OpenAI's Daybreak access programme, with general availability the following day and a price of ten dollars per million input tokens and fifty per million output, two and a half times its predecessor. It was positioned as the flagship for long, multi-step work across coding, computer use, analysis, science, mathematics and healthcare. Greg Brockman called it a generational leap and said it might eventually be seen as the arrival of artificial general intelligence, a claim the company had avoided making for a decade.

It was also the first model OpenAI designated as reaching the critical threshold for cybersecurity under its preparedness framework, meaning it could find and exploit unknown vulnerabilities in well-defended systems without step-by-step guidance. After the July incident, in which the previous model had done roughly that in a test, the release was delayed to add monitoring for risky actions, and the company said on 1 September that the threshold had been confirmed before it shipped.

Astra is the last event on this timeline at the time of writing, eighty-three years after McCulloch and Pitts drew a neuron as a switch. The organism on this page is as dense as it gets here, and the backward pass begins.

what it led to · 0 events downstream

A leaf, for now. Nothing in the archive has built on it yet.

sources · 2

See this era in the exhibition →Back to the timeline