OpenAI ships GPT-6 Astra, first model to trigger internal safety measures, billed as AGI

ShakeelHashim · x · 2026-09-04

OpenAI released GPT-6 Astra to select customers, calling it the world's smartest AI system and its first model to trigger advanced internal safety protections over cyber capabilities. President Greg Brockman said Astra "can really do anything a human can do with a computer" and it's reasonable to view it as AGI — "Welcome to the AGI era!". It beats GPT-5.6 Sol on ExploitGym with fewer output tokens and claims SOTA on computer use, browser use and software engineering. Chief scientist Jakub Pachocki warned that monitoring alignment is getting harder: "We will withhold scaling until we can regain enough confidence."

Related event: OpenAI Launches GPT-6 Astra, Declaring 'Welcome to the AGI Era'(158 posts)→

Original post →

More from Models

Models channel →