Starting in December, your ChatGPT and Codex usage will come out of the same allowance. OpenAI updated its help center over the weekend: Plus, Pro and Business plans will get one shared pool, measured in tokens, as the company folds the agentic tools of Work into the main ChatGPT app.
Work is ChatGPT's slow mode, the one that goes off and runs a long research job or a multi-step task while you wait. Codex is OpenAI's coding agent. Until now they've had their own limits. From December, a long research run in ChatGPT and a big refactor in Codex draw on the same budget.
There's one exception, and for many people it's the whole story. OpenAI writes: "Instant chat will not draw from your usage." If you mostly ask ChatGPT quick questions on the default fast model, December changes nothing for you.
Who should pay attention
You'll feel this if you use ChatGPT's heavier models for real work and also code with Codex. Token pricing means the bill depends on what you ask for. A Thinking-model answer that reads twenty sources costs more than a two-line reply, and OpenAI says more advanced models and complex tasks "generally use more of your allowance." We saw the same pattern with Anthropic's Sonnet 5.5, where the effort setting moved the cost more than the price card did.
What OpenAI hasn't said yet
The pushback started fast. Threads on OpenAI's developer forum ask the company to keep the limits separate, and a GitHub issue filed by a Pro subscriber warns that the shared quota risks "a major effective price increase." That's one user's reading, and OpenAI hasn't published how many tokens each plan actually gets. Its own line is calmer: "We expect most users won't need to wait for their usage to reset more often than they do now."
Maybe. But "most users" is doing a lot of work in that sentence, and the people who'll hit the wall first are exactly the ones paying for higher plans.
What to do before December
Get a feel for which of your habits are expensive. When the change lands, OpenAI says your combined remaining usage and reset times will show up under Settings, then Usage. Check it after a normal working day in the first week of December, before you start a big job.
Keep quick lookups on Instant, since those stay free of the meter. Save the Thinking models and Work for tasks that actually need them, like a long report or a messy spreadsheet. If you code a lot, decide whether Codex really belongs on your chat subscription. Codex can also run on a pay-as-you-go API key, which keeps your coding from eating the allowance you need for everything else.
If you do run out, OpenAI lists three options: wait for the reset, upgrade your plan, or buy credits. We'd treat the first week as a test run and only upgrade once you've seen your own numbers.
The meter has always been there. In December you'll finally see it as one number.
Cover: AI-generated editorial illustration by TMRW using OpenAI branding. The meter and cables are illustrative.
