News · 2026-10-08

ChatGPT Pro’s $200 tier now includes a lower allowance, with an October transition for existing users

OpenAI lowered the included usage allowance for new, non-grandfathered subscribers to its $200 ChatGPT Pro tier while keeping the monthly price unchanged. Eligible existing subscribers retain the previous allowance through October 29 before moving to the lower level afterward. The change affects Codex users, but official documentation does not establish that every subscriber can complete exactly half as much coding work.

Key facts

• The monthly Pro 200 price remains $200 despite a lower included allowance for affected subscribers.
• OpenAI announced reopening the tier and changing usage in its September 29 release notes.
• Eligible existing subscriptions retain the former allowance through October 29 if active.
• The primary policy source is OpenAI’s Pro-tier help page.

The story has renewed attention because heavy coding-agent users experience subscriptions through work completed, rather than an abstract usage balance. An allowance that once felt difficult to exhaust can become restrictive as agents make more calls, use more context or run stronger models. That experience can change even if the account’s monthly charge stays the same.

The September 29 release notes announce that Pro 200 became available again and its included usage was changing. OpenAI’s detailed help page gives grandfathering conditions: an active subscription during the specified September 22–29 window preserves the earlier allowance through October 29, provided the subscription stays active. New customers outside that eligibility receive the lower allowance. The policy is a subscription change, rather than an API price change.

The viral “half” wording needs its source attached. Codex lead Thibault “Tibo” Sottiaux said the change would “net out at half the dollar in API spend compared to the old Pro $200 plan.” His September 29 post is reproduced in the OpenAI Developer Community. The direct social-media page was inaccessible in the research pass. The reproduced statement gives an accounting comparison, not a published conversion into prompts, tokens, sessions or completed projects.

OpenAI’s Codex pricing guide explicitly distinguishes API token pricing from included subscription usage. Work and Codex share usage. Consumption depends on the model, task size, reasoning, context, tools, retrieval, caching, execution location and speed. Two requests that look equally short to the user can therefore draw different amounts from the allowance.

A useful analogy is a travel pass that covers several kinds of journey. The price of the pass is one quantity. Its included travel balance is another. How many trips the holder finishes depends on the routes and services selected. Knowing an equivalent retail value can explain the provider’s accounting, but it does not tell every traveler how many journeys remain. A coding task likewise involves a changing combination of model calls and tool activity.

The distinction gets sharper with speed settings. OpenAI’s current guide says Fast mode consumes included usage at 2.5 times the Standard rate, while Astra Ultrafast consumes it at eight times. The Pro-tier help page limits Ultrafast to Pro 500. Those settings are not interchangeable workloads, and comparing a fast run with a standard run without recording the setting obscures the effective plan value.

The strongest operator reaction in the dossier is Theo’s October 7 video. He argues that earlier subscriptions felt practically unlimited and that revised accounting changes their value for heavy users. His API-equivalent totals and monthly extrapolations describe his own workflow. They are useful testimony about one intensive use pattern, but they do not establish the general plan budget or the median subscriber’s experience.

An Anthropic comparison in the discussion also needs tier-specific treatment. The Fable help page says eligible Max and specified premium seats can use Fable for up to half of their existing weekly allowance. That is a share of the same pool, not an additional half-plan. Pro and standard seats generally use credits instead under the documented terms. A broad statement about half of every Claude subscription would conceal those differences.

The banked-reset Reddit discussion captures mixed reactions to plan value and reset timing. Its title and comments do not establish the mechanics of a future reset feature. The reviewed official policy does not explain how such a feature would interact with weekly windows or credits, so those details remain outside this report.

The practical consequence is a more explicit budget for open-ended agent work. Readers can connect it to the existing hard-budget story and the lesson on inference economics. The honest caveat is that official pages do not publish a universal numerical task allowance. A lower included balance is verified; an exact reduction in every person’s useful work is not.

Primary source, verified: read the paper →

Ground Truth. (2026, October 8). ChatGPT Pro’s $200 tier now includes a lower allowance, with an October transition for existing users. Ground Truth. https://groundtruth.day/news/chatgpt-pro-200-lower-codex-allowance.html

@misc{groundtruth:chatgpt-pro-200-lower-codex-allowance,
  title  = {ChatGPT Pro’s $200 tier now includes a lower allowance, with an October transition for existing users},
  author = {{Ground Truth}},
  year   = {2026},
  month  = {oct},
  url    = {https://groundtruth.day/news/chatgpt-pro-200-lower-codex-allowance.html}
}

Topics: openai · codex · subscriptions · pricing · developer-tools