OpenAI has temporarily stopped new sign-ups for its highest-usage consumer subscription, after demand for its latest model, Astra, outpaced the company's capacity.
The pause applies to new subscriptions and upgrades for the $200-a-month Pro plan.
What still works, and what doesn't
Existing Pro accounts continue to function normally, and other subscription plans remain on sale.
OpenAI's product leader, Thibault "Tibo" Sottiaux, wrote on X that demand for Astra is "really unprecedented" and that the Pro tier "puts the most strain on our systems."
Sottiaux added that the company was pulling "all the levers possible to sustain the demand."
Lower-cost consumer tiers and OpenAI's API both remain available, a sign the restriction is targeted specifically at the most compute-intensive subscription, rather than reflecting a platform-wide outage.
Why one model caused this surge
Astra is OpenAI's newest flagship model, and has been rolled out across the company's paid offerings since its launch on 3 September, appearing on both consumer and business plans.
OpenAI has touted Astra as a significant step forward for reasoning, coding and computer control.
That positioning helps explain the surge: a flagship model that both promises new capabilities and sits inside higher-priced accounts concentrates unusually heavy compute demand into a relatively small part of OpenAI's overall billing structure.
An operational fix, not a product change
OpenAI has not published user numbers or a timeline for reopening Pro sign-ups, underlining that this is intended as a short-term operational response rather than any change to the product itself.
The company says it is adding capacity and may reopen Pro subscriptions once it can guarantee service for existing users.
What it means for users, and for the industry
For heavy users and teams relying on the top-tier performance Astra promises, the immediate consequence is straightforward: reduced access until OpenAI expands capacity.
For the wider market, the episode serves as a reminder that flagship model launches now create supply problems just as often as they create demand ones, testing whether AI companies can scale backend infrastructure quickly enough to match the spikes their own product releases generate.
