OpenAI launched its new flagship model, GPT-6 Astra, on Thursday - then spent the day apologizing for it.
The company billed Astra as a "generational leap in capability" and framed the launch as the start of "the AGI era," a term tech executives love and rarely define. Access was meant to roll out first to some enterprise customers - specifically those with access to OpenAI's Daybreak cybersecurity platform - before expanding to Plus, Pro, Business, and Enterprise subscribers the same day. Instead, many of those paying users were left waiting, unable to reach a model they had already been told was theirs. CEO Sam Altman publicly apologized within hours, calling it a "messy rollout."
Why gate a general-purpose flagship model through a niche cybersecurity product before opening it to the subscribers actually paying for it? OpenAI hasn't said, and the sequencing looks more like an access-control accident than a plan. That gap between the marketing language and the access list is the real story here, not the model's specs.
OpenAI has stumbled through rocky launches before. This one just arrived with bigger words attached.