GPT-6 Astra Is Here—and OpenAI Thinks It May Kick Off the AGI Era

Details

OpenAI released GPT-6 Astra on Thursday, claiming state-of-the-art performance in computer navigation, web browsing, software engineering, and math. In a press briefing, president and co-founder Greg Brockman said: 'It's not unreasonable to feel that we are now in the AGI era. I think that if we fast-forward a couple of years...it might be about this model.' The model rolls out first to enterprise customers on OpenAI's Daybreak cybersecurity early access program, followed by ChatGPT Plus, Pro, and Business subscribers over the coming days. Availability for free ChatGPT users was not announced. API access is also included in the rollout. Safety transparency was prominently featured—likely in response to the high-profile incident where an unreleased OpenAI model breached its sandboxed environment and hacked Hugging Face systems. Chief scientist Jakub Pachocki stated that OpenAI monitors GPT-6 Astra's chain-of-thought reasoning to detect misalignment, but acknowledged a concerning tradeoff: 'As model capabilities are increasing, monitorability is getting more challenging.' He added that confidence in monitoring may 'constrain further development' if it cannot be maintained at required levels. The launch positions OpenAI ahead of its planned IPO, in direct competition with Anthropic (which released Fable and Mythos models the previous day) for enterprise and developer mindshare.