ChatGPT and Codex limits to be shared from December
OpenAI said on a help page on Oct. 31 (Nov. 1 Japan time) that it will revise usage limits for paid plans of its conversational AI "ChatGPT." Starting in December, the three plans, Plus, Pro and Business, will count usage of ChatGPT and the coding assistant "Codex" against a single shared allowance. The allowance is counted not by the number of uses but in tokens, the units of text that AI processes. OpenAI did not give the start date or the size of each plan's allowance.
There was no announcement on the official blog or on X, only an update to the help page. The update was found and reported by "LuminaBench," an X account that tracks unannounced AI information. The X post reporting the change was viewed more than 89,000 times in six hours.
On Plus and Pro, Codex, "ChatGPT Work," a feature for delegating tasks, and the Excel and PowerPoint features already share an allowance for uses in which work is delegated to AI. With this change, conversations in ChatGPT will also come under this allowance. OpenAI explained that the change is in line with its move to fold ChatGPT Work's features into ChatGPT itself.
More capable models use the allowance faster
The help page explains that more capable models and more complex tasks use more of the allowance. Conversations using the standard conversational mode "Instant" do not use the allowance. When the allowance is used up, users can choose from three options: wait for it to reset, switch to a higher plan, or buy credits for the covered features. The remaining amount and reset time can be checked under "Usage" in settings.
The difference by model is large. In the rate table OpenAI publishes for the Excel and PowerPoint features, credits per 1 million tokens are 5 for input and 30 for output for the low-cost "GPT-5.6 Luna," and 250 for input and 1,250 for output for the top-end "GPT-6 Astra." This rate table is separate from the conversion for the shared allowance, and OpenAI has not disclosed how tokens will be counted under the shared allowance.
OpenAI expects that most users will not have to wait for resets more often than they do now.
Allowance sizes and start date not disclosed
Whether users gain or lose cannot be determined until the allowance sizes are released. The size of each plan's allowance, the start date, and how ChatGPT and Codex usage will be converted under the same allowance have all not been disclosed. An outside commentary described it as simpler, but not more generous.
A single allowance offers more flexibility in how it is used. In a month when Codex is used little, the remainder can go toward conversations in ChatGPT. On Business, two balances become one, making it easier for teams to manage.
On the other hand, continuing large development work in Codex will also reduce the ChatGPT allowance for the same period. Plus and Business have five-hour limits, and sharing the allowance with Codex could mean hitting these limits sooner. How Pro's weekly limits relate to the shared allowance has also not been explained. That users will not have to wait for resets more often is OpenAI's expectation, and it may not hold for heavy users.
In October, a cut to limits was reversed and limits doubled
OpenAI's limits changed twice in October alone. At its developer conference at the end of September, OpenAI announced that from Oct. 30 it would reduce ChatGPT Work and Codex usage on the $200-a-month "Pro 200" from 20 times Plus to 10 times. At the same time, it created "Pro 500," at $500 a month, offering 25 times Plus.
On Oct. 29 (Oct. 30 Japan time), the day before the cut was to take effect, Codex lead Thibault Sottiaux said on X that limits for Pro 100, Pro 200, Pro 500 and the business-oriented Business Premium would be doubled, and the Pro 200 cut was also reversed. Pro 500 usage became 50 times Plus.
On the same day, users found a note on ChatGPT's purchase screen saying that starting in December, plan usage would be a shared limit across ChatGPT and Codex.
- Announces cut in Pro 200 usage from 20 times Plus to 10 times. Pro 500 createdOfficial
- Limits for all Pro tiers and Business Premium doubled; Pro 200 cut reversedOfficial
- Note on shared limit from December found on purchase screenLeak
- Anthropic renames Max 20x as Ultra and removes the five-hour limitOfficial
- OpenAI specifies the shared allowance on a help pageOfficial
OpenAI has not explained whether the doubled limits will carry over directly into the shared allowance from December.
Anthropic removes five-hour limit in the same week
Rival Anthropic on Oct. 30 (Oct. 31 Japan time) changed the names and limits of paid plans for its conversational AI "Claude." It renamed "Max 5x" as "Max" and "Max 20x" as "Ultra," keeping prices unchanged at $100 and $200 a month. Ultra users who have completed identity verification no longer have the five-hour limit. Weekly limits are unchanged, and Ultra remains at 10 times the lower-tier Pro.
There is no application process for identity verification; Anthropic carries it out automatically based on account and payment history. The five-hour limit remains until verification is complete.
Anthropic has also been criticized over how it presents limits. On Sept. 8, users sued Anthropic, alleging that Max's labels of 5 times and 20 times Pro compared only the five-hour allowance and did not adequately disclose the weekly limits.
Moving closer to counting by usage
ChatGPT has more than 1.2 billion weekly users, and ChatGPT Work and Codex together have 35 million weekly users. Paid users numbered about 50 million as of February 2026.
With this change, limits on paid plans will be counted by the amount of tokens used, with any shortfall purchased with credits. Business seats contracted on a credit basis already pay for ChatGPT Work and Codex usage at per-token rates. Pro and Business separately come with up to $500 a month in developer API credits.
