BUSINESSNew tool · · 2 min read · Source: OpenAI Help Center
ChatGPT: Chat, Work and Codex will share one limit from December
OpenAI will move ChatGPT and Codex to one shared usage allowance starting in December 2026. The change covers the Plus, Pro and Business plans and appears on a help page updated on October 31.
Key facts
- From December 2026, ChatGPT and Codex will share one allowance on Plus, Pro and Business; no exact day has been given.
- Usage is measured in tokens; simple Instant chats do not count against the allowance.
- At the limit, users can wait, upgrade or buy credits; OpenAI has not yet published figures per plan.
What is changing
Until now, Work and Codex have drawn on their own allowance, while regular chat has had separate limits. The new setup uses one shared allowance, measured in tokens. Put simply, tokens are the unit that counts how much text a model processes.
- More advanced models and complex tasks use up more of the allowance.
- Simple chats with the Instant models do not count.
- Your remaining budget and reset time are due to appear in settings.
What happens at the limit
According to reports on the help page, you will have three options: wait for the reset, move to a higher plan, or buy extra credits. ChatGPT stays a subscription, but demanding work now visibly eats into a budget. A long agent job on Monday could affect what you can do for the rest of the week. The same reports say OpenAI expects most users not to wait for resets more often than they do today. That is an expectation, not a promise.
What is still unknown
OpenAI has not yet shared several important details:
- the exact start date in December,
- the size of the allowance per plan,
- how today's time windows will carry over to the shared pool,
- what happens to the Free, Go, Enterprise and Edu plans.
The move fits a wider rebuild. In October, Tibo Sottiaux, who oversees ChatGPT and Codex at OpenAI, said Chat and Work would become one experience, without giving a firm date. Anthropic has already merged chat and Cowork in Claude.
What this means for you
- Heavy users of ChatGPT with Work or Codex will share one budget from December, so big jobs can leave less room for other work.
- Model choice becomes a cost decision: not every task needs the strongest model.
- Business teams should map who uses ChatGPT for what now, so they can spot early whether the allowance will be enough in December.
How to use this
By the end of November, note which tasks really need Work or Codex and which a faster model can handle. In December, the settings panel should then quickly show whether your plan still fits.
Luigi's take
This is the second story today pointing the same way: Microsoft is also billing top models in Copilot by usage. AI subscriptions still look like flat rates, but they increasingly behave like budgets. That is no scandal, because agents that run for a long time cost providers a lot of computing power. But if your team builds its processes on one tool, you should know what happens when the budget runs out on a Wednesday.
My advice: do not wait for the switch. Sort your recurring tasks into two groups: what the fast model can handle, and what truly needs the strong one. Until OpenAI publishes numbers per plan, I would not base an annual contract or a critical process on this alone. See the numbers first, then decide.
Source
OpenAI Help Center: Understanding ChatGPT and Codex usage limits
More sources
- The Decoder – Bericht zur Änderung, Einordnung und Vorgeschichte
This article was researched and written with AI assistance. Responsible: Luigi Lugmayr.
