Thibault Sottiaux, OpenAI's product lead for Codex and ChatGPT, said that demand for the GPT-6 Astra model has broken all of the company's internal records, and OpenAI may temporarily stop new ChatGPT Pro signups to maintain service quality.

image

What happened

On September 9, 2026, Thibault Sottiaux posted on X that load on GPT-6 Astra has exceeded all of OpenAI's internal historical metrics, and the company is using all available compute capacity to keep infrastructure available. According to him, if load growth continues, OpenAI may temporarily pause new ChatGPT Pro signups: this is a preventive measure to avoid inference queues and performance degradation. The company named maintaining service quality for current subscribers as its priority. Independent retellings of the original message appeared on September 9 in PANews and nnets.ru.

Context

GPT-6 Astra launched on September 3, 2026, and the demand record was set about six days after release. The phrasing that demand broke all internal historical records is an OpenAI operational metric without a public methodology or quantitative breakdown. Sottiaux's statement contains no technical details about the model: no benchmarks, no comparisons with previous generations, no architecture details, no training methods. Thus, the news is operational rather than scientific: it describes infrastructure load, not the capabilities of GPT-6 Astra.

Why this matters for the industry

For the LLM inference market, this is a rare public example of a capacity crunch: user demand is physically hitting the throughput of inference infrastructure, and the limiting factor is not request price but model availability itself. OpenAI's willingness to temporarily forgo new revenue from ChatGPT Pro sales for service stability for existing clients shows that capacity management, capacity planning, and capacity-aware UX are becoming independent competitive factors. For teams integrating models into production, this is a signal: without a fallback model, critical scenarios risk increased latency and failures under peak load.

Why this matters for users

Those planning to switch to ChatGPT Pro should consider that new signups may be temporarily paused. Current subscribers will retain access. After launch, GPT-6 Astra became OpenAI's most in-demand model, so in the coming weeks it is likely to have queues, limits, and uneven response quality during peak load hours.

What is still unknown / limitations

The demand record is not a benchmark and does not prove the technical superiority of GPT-6 Astra: it may be inflated by launch effect, marketing, and exclusive distribution. There is no quantitative breakdown of the record or public methodology for evaluating it, so this statement cannot be independently verified. The company has not announced exact timelines and conditions for a possible pause of new ChatGPT Pro signups.

Sources

Author

Look at AI, editorial team