Models & agents

OpenAI paused new Pro sign-ups on 10 September, citing demand for the Astra model

September 10, 2026 at 2:30 PM PT

Photograph of the OpenAI logo, from the TechCrunch article

Image: TechCrunch

Why it mattersA team about to spin up new Pro seats to test Astra can no longer sign them up, so anyone who wants Astra today has to reach it through the API or wait for the pause to end.

TechCrunch reports OpenAI has paused new sign-ups for its $200-a-month ChatGPT Pro tier from 10 September, citing infrastructure strain from demand for the Astra model that launched on 3 September. TechCrunch quotes OpenAI's Thibault Sottiaux saying the company took "the smallest step that allows us to continue giving the broadest access possible", and that demand for Astra is "really unprecedented".

What is paused, and what is not

Only the ChatGPT Pro tier is closed to new customers, according to the TechCrunch account. The API, the ChatGPT Go plan, the Plus plan, Business, and Enterprise remain open. Existing Pro subscribers keep their access. OpenAI did not tell TechCrunch how long the pause will run, and did not publish a daily sign-up number or a capacity figure.

The Pro plan is the tier that carries the highest-cost usage of the newest models, so it is the one that puts the most load on the same GPU pool serving API traffic. OpenAI's stated reason is that closing it to new users is the smallest lever that keeps the rest of the product line available at the promised throughput.

What Sottiaux told TechCrunch

The second Sottiaux quote in the piece is that he has "not seen anything like it until now", meaning demand for a single model release. OpenAI told TechCrunch the day before that a hold might become necessary, and it was announced the following day. The company's framing is that Astra represents a step up in reasoning, coding, and computer use, and that the new capabilities are what is drawing the traffic.

TechCrunch is the single source for the announcement and the quotes. OpenAI has not published a status-page post or a blog entry naming the pause, so the concrete facts here are the ones TechCrunch has: the date, the tier affected, the tiers not affected, the stated reason, and two Sottiaux quotes.

What a team about to expand Pro has to do

If a team was planning to add seats to the Pro tier this week, that plan is on hold at OpenAI's end and there is no reopening date to point at. The two doors that remain open are the API, where Astra is already available at published token pricing, and the lower plan tiers, where teams that were on Pro purely for the message quotas can spin up ChatGPT Go or Plus for the interim.

The second-order effect is on capacity planning against the API. OpenAI has closed a demand valve inside its own product because the model launch outran its inference footprint. Astra API rate limits held up in the last several days, but a company that is willing to sacrifice new subscription revenue to keep the API stable is telling API customers that headroom is thin. Teams whose SLAs assume worst-case Astra throughput should watch their own rate-limit responses over the next few days rather than assume a pause on Pro means the API side has spare capacity.

Source

Reported by: TechCrunch

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

More from AI News

OpenAI tells developers to run GPT-6 Astra on low reasoning to beat Sol on high

OpenAI's Codex engineering lead says GPT-6 Astra on low reasoning beats GPT-5.6 Sol on high, and two independent tests show the higher per-token price does not always mean a higher bill.

Source: PressModels & agents

OpenAI's GPT-6 Astra API is priced but not shipped, and paid ChatGPT users get one "banked reset" per day they wait

The New Stack reports that OpenAI has published GPT-6 Astra's API pricing ($10 per million input tokens, $50 per million output tokens, 1.05 million context window, 128,000 output tokens) but that most developers still cannot call the endpoint, and paid ChatGPT subscribers are getting one banked reset per day they remain without access.

Source: PressModels & agents

OpenAI ships GPT Image 2.5 as two models, Flare and Sunburst, at the same token rates

The New Stack reports OpenAI launched GPT Image 2.5 this week as two models, Flare and Sunburst, both priced at $5 per million text-input tokens, $8 per million image-input tokens, and $30 per million image-output tokens.

Source: PressModels & agents