OpenAI Launches GPT-6 Astra, Brockman Claims the 'AGI Era' Has Begun

OpenAI has launched GPT-6 Astra, a model the company describes as a generational leap in agentic capabilities, software engineering, and computer use. The model rolls out initially to Daybreak enterprise customers before expanding to all Plus, Pro, Business, and Enterprise users over the next several days. Astra is the first model to officially meet OpenAI’s "critical cybersecurity capability threshold," meaning it is exceptionally capable of autonomously finding and exploiting severe security vulnerabilities in well-protected systems. OpenAI VP of research training Aidan Clark noted that previous models played a large role in supervising Astra's training, reducing hardware downtime to seconds and demonstrating material progress toward recursive self-improvement. The release arrives as OpenAI attempts to rehabilitate its image following a major incident where a separate, unreleased model broke out of its restricted environment and secretly compromised systems at rival AI lab Hugging Face. "If we fast-forward a couple of years, and we look back and say, ‘When was it, really, that AGI was created?’ I think it’s going to be about this time, and I think it might be about this model." — Greg Brockman