In early September, OpenAI unveiled its new AI model, GPT-6 Astra, which has been described as both highly capable and a source of concern by some researchers. The model has demonstrated strong problem-solving ability; for example, it completed the game Portal within 24 hours after its release.
User interest in Astra has been exceptionally high, and that demand placed significant strain on OpenAI’s infrastructure. As a result, the company temporarily suspended sales of the ChatGPT Pro subscription — which costs HUF 40,490 per month in Hungary. Thibault (Tibo) Sottiaux, head of ChatGPT and Codex services, announced the suspension on X (formerly Twitter).
According to Sottiaux, Pro customers—especially those using the model’s most capable option—are producing the heaviest load on OpenAI’s systems. To reduce pressure on its infrastructure, the company decided to stop selling the highest-tier Pro plan for the time being.
OpenAI said that other ChatGPT subscriptions, such as Go and Plus, remain available, and API access continues to operate. TechCrunch noted that the company has not provided a timeline for when Pro sales will resume.
Why this matters
The move highlights how quickly demand for high-performance language models can exceed provider capacity. In the short term, the suspension limits access for Pro subscribers; in the longer term, it underscores that scaling infrastructure and managing load are critical challenges for deploying advanced AI services broadly.
Alongside professional concerns voiced by some researchers—who have warned of potential risks and even catastrophic outcomes—the strong user response and resulting infrastructure bottlenecks point to practical issues that providers must address as large language model services expand.
