Published: 2 November 2026
From December, ChatGPT Plus, Pro and Business subscribers will spend one usage allowance on both ChatGPT and Codex, OpenAI’s coding agent. OpenAI set out the change in an updated help centre article at the end of October. Chats with Instant models stay outside the allowance. Chats with more advanced models, Codex tasks and agent work in ChatGPT all draw from the same budget.
A long Codex session will leave less room for chats with advanced models
Codex and Work, the agent mode in ChatGPT, already share one allowance. Ordinary chats have had their own limits. In December those two pools become one. OpenAI’s help article puts it plainly: “Starting in December, Plus, Pro, and Business plans’ usage allowance will be shared across ChatGPT and Codex.”
The allowance is measured in tokens, which OpenAI describes as “the units of information a model processes and generates”. More advanced models and complex tasks use more of it. A developer who keeps Codex busy on a large refactoring job therefore spends the budget they would otherwise use for research or document analysis in ChatGPT. Quick questions to an Instant model cost nothing from the allowance.
The change is part of OpenAI’s plan to bring Work into the main ChatGPT screen. In early October, Tibo Sottiaux, who runs ChatGPT and Codex at OpenAI, said Chat and Work would merge into one experience, according to Startup Fortune.
Settings will show one combined balance and the time it resets
Under Settings and then Usage, you will see the combined balance and its reset time. When it runs out, you can wait for the reset, move to a bigger plan or buy credits to keep using eligible features.
Plus costs $20 a month. Pro comes at $100, $200 or $500 a month. Business costs $25 per user per month, or $20 when billed annually, and Business workspaces with flexible pricing can buy extra workspace credits.
Credits are where your model choice shows. On OpenAI’s credit rate card, a million output tokens from GPT-6 Astra costs 1,250 credits, against 12.5 credits for GPT-6 Luna. OpenAI has not said whether the shared allowance is counted at the same rates.
OpenAI has not published how large the shared allowance is
The help article gives no allowance per plan, no reset periods and no conversion from chat tokens to allowance. Today, OpenAI’s pricing page gives Plus and Standard Business seats a five-hour usage window, with possible weekly limits on top. Pro has no five-hour limit. Whether that structure stays after December is not stated.
OpenAI does offer a reassurance: “We expect most users won’t need to wait for their usage to reset more often than they do now.” It publishes no figures to back that expectation. Subscribers on OpenAI’s developer forum have asked the company for the included capacity and the reset periods.
Does the shared limit apply to ChatGPT Enterprise and Edu?
The help article does not say so: it names only Plus, Pro and Business. Free, Go, Enterprise and Edu are not mentioned. Enterprise and Edu workspaces on flexible pricing already pay per credit and have no fixed rate limits, according to OpenAI’s pricing page. OpenAI mentions no regional exceptions, so subscribers in the Netherlands and the rest of Europe should expect the switch in December as well.
Business admins should know their heaviest Codex users before December
A Business seat costs the same for every colleague, but use rarely is equal. A few developers and analysts run long agent tasks, while most people ask short questions. One shared allowance turns that difference into a question of who gets stuck first. For the organisations DataNorth works with, November is the month to find out who leans on Codex and Work. Agree in advance who gets a Pro seat or credits when the allowance runs out, so a project does not stall halfway through a sprint.
Point routine questions to an Instant model, which does not touch the allowance, and save the advanced models for work that needs them. Do not buy upgrades on a guess: wait for OpenAI’s numbers, then compare the first week of December with your team’s actual usage.
For more information, visit the official explanation of the shared usage limit in the OpenAI Help Center.
