Morning Edition · Friday, September 25, 2026Published at 9:31 AM EDT · New York
The unannounced plan would cost 2.5 times the current $200 Pro tier, with the upgrade described in the client as "Fastest Work and Codex."

Strings found in the ChatGPT client point to an unannounced subscription called Pro Max, priced at $500 a month, first reported by TestingCatalog on September 24. The described benefit is higher usage limits and "Fastest Work and Codex," which suggests guaranteed throughput on long-running agentic and coding jobs rather than access to a new model. OpenAI has not confirmed the tier, and it does not appear on the pricing page. The company's DevDay is scheduled for September 29 in San Francisco.
Russian-language AI channels tie the move directly to serving capacity. The channel AI ML Big Data wrote that the step follows a serious compute shortage after the launch of GPT-6 Astra, which consumes far more resources per request than its predecessors. That causal claim comes from the channel, not from OpenAI. An English-language summary described the same leak without the capacity framing.
What is documented is that Astra access is metered tightly. OpenAI publishes estimated Astra allowances for Work and Codex messages per five-hour window, ranging from single digits on the Plus plan to several hundred on the $200 Pro plan, and states that these are estimates rather than fixed caps. A $500 tier would effectively price priority scheduling for customers whose agents run for hours, a different product from a standard chat subscription.
OpenAI, if the leak tests price tolerance before DevDay without commitment, and accelerator and inference-hardware suppliers, for whom visible rationing of Astra capacity is evidence of demand exceeding supply.
The code strings and the $500 figure are documented by TestingCatalog and researcher Tibor Blaho, but OpenAI has confirmed nothing, some coverage shows a $600 checkout figure that reflects value-added tax rather than the base price, and the causal claim that the tier follows an Astra compute shortage comes from Telegram commentary, not from the company.
Start a discussion in Townsquare.
More from this edition
An open-source-intelligence read of how likely this story is true with its real nuance, not a judgment of any outlet. It assesses the claim, weighing independent and adversarial reporting. How we label confidence.
What this means
The mechanism is scarcity rationing through price discrimination. When reasoning-model serving capacity is limited, the more profitable move is to sell scheduling priority to the users most willing to pay for it, not to cut prices. That splits the market: Anthropic is cutting posted API rates on its flagship model, while OpenAI appears to be building a premium tier above $200, and both moves can be rational at once if one company currently has more spare serving capacity than the other. Independent coding-agent startups that resell Codex-style capacity are exposed, since they cannot pass a $500 price floor on to their own users, and accelerator suppliers benefit, since sustained rationing is evidence of demand. The precedent is reserved-instance pricing at cloud vendors, where guaranteed capacity has always commanded a premium over spot access. The likelier outcome is confirmation of a top tier at or near this price at DevDay on September 29, with per-window limits rather than unlimited use, conditional on Astra's serving costs staying where they are. Silence at DevDay, or the release of a cheaper Astra variant instead, would indicate the capacity crunch is easing faster than the leak suggests.
What to watch
Observations to monitor, not financial advice.
Synthesized from: Polylog editors · TestingCatalog
Comments
0No comments yet.