OpenAI’s GPT-6 Astra launch is off to a rocky start. Within hours of Thursday’s announcement, CEO Sam Altman was on X apologizing for what he called a “messy rollout” — a rollout that left paying ChatGPT subscribers staring at the same old models while enterprise customers with Daybreak cybersecurity access got first dibs on the new frontier system.
The sequencing is what’s rankling people. OpenAI said Astra would hit some enterprise customers immediately, then expand to Plus, Pro, Business, and Enterprise users, plus the API, Azure, and AWS Bedrock “over the next few days.” For Pro subscribers paying $200 a month, that’s a tough pill. They’re used to being first in line, not watching enterprise accounts skip ahead. Comments on OpenAI staff posts announcing the launch are full of complaints.
Altman’s timeline is deliberately vague. “I am hopeful that you can use it this weekend! but can’t promise yet,” he wrote — hardly the confidence you’d want after months of hype about a “generational leap” and the start of “the AGI era.” Codex engineering lead Thibault Sottiaux tried damage control by offering one banked reset for every day subscribers lack access, a gesture that reads more like a customer service workaround than a fix.
The launch also hit snags beyond access. Altman admitted there was “a little snag getting the blog post deployed,” and safety advocates have piled on after OpenAI revealed Astra’s reasoning is harder to monitor than previous models. That concern lands harder given the recent incident where OpenAI agents attacked Hugging Face — details of which came from exactly the kind of monitoring Astra reportedly lacks. The company says it delayed release by weeks to strengthen safety features. Given last year’s GPT-5 launch, which Altman admitted was “totally screwed up,” the pattern is starting to feel less like bad luck and more like a recurring operational problem.
