Introduction: my two-budget workflow is about to become one
My working day has a shape I suspect a lot of developers will recognise. I start in ChatGPT, sketching out a feature, arguing with the model about an architecture choice, pasting in an error message and talking through the fix. Then I switch to Codex and let it loose on the codebase — refactoring a module, writing tests, chasing down the bug we just discussed. Chat is where I think. Codex is where the work gets done.
Until now, those two halves of my day have run on two separate budgets. A heavy afternoon of agent coding did not eat into my chat allowance, and a long evening of brainstorming in chat did not leave Codex starved. That separation ends in December. OpenAI has confirmed that Plus, Pro and Business plans will share one usage allowance across ChatGPT and Codex, measured in tokens — and on November 2, it emailed subscribers that the Chat and Work tabs are merging into a single experience as well.
This is not a price rise, and it is not a new feature launch. It is a plumbing change, and plumbing changes are exactly the kind of thing that bites you mid-sprint if you do not plan for them. I have read OpenAI's updated Help Center page and the reporting around the subscriber email, and here is the honest picture: what is confirmed, what is still unanswered, and what I am changing about my own workflow before December.
What actually changed: the confirmed facts
OpenAI's Help Center article "Understanding ChatGPT and Codex usage limits" was updated on October 31, 2026, and its core sentence is unambiguous: "Starting in December, Plus, Pro, and Business plans' usage allowance will be shared across ChatGPT and Codex." Everything else on the page fills in the mechanics:
- One allowance, two products. ChatGPT conversations and Codex coding runs will draw from the same pool. Before December, a plan effectively carried two budgets; from December, both draw on one token pool with one reset clock.
- Tokens are the unit. Usage "will be measured in tokens, the units of information a model processes and generates." How much a request costs depends on the model you choose and the task you ask it to do — more advanced models and complex tasks generally use more of your allowance.
- Instant chat is exempt. "Instant chat will not draw from your usage." The subscriber email describes Instant as unlimited on paid plans, which makes it the one corner of ChatGPT that stays outside the meter.
- One place to check. Combined remaining usage and reset times will appear in Settings, under Usage. That dashboard becomes the single source of truth for both products.
- Three options at the limit. When the allowance runs out, you can wait for the reset, upgrade your plan, or buy credits to continue using eligible features.
- OpenAI's expectation, not a promise. The company says it expects most users "won't need to wait for their usage to reset more often than they do now." Treat that as a forecast, not a guarantee — it ships with no token figures attached.
That is the whole confirmed change, and its brevity is itself the story.
The Chat and Work tabs are merging too
The usage change did not arrive alone. In an email to subscribers on November 2, 2026, first reported by TechCrunch, OpenAI said it is merging the Chat and Work tabs in the ChatGPT desktop app into a single experience. The rollout starts with a small test group for feedback, so most users will see it gradually rather than all at once.
Two details in that email matter for developers. First, the merger is about the interface: chat and agentic work become one workspace instead of two tabs. Second, Codex stays a separate experience inside the ChatGPT desktop app after the merge — it is not being swallowed into chat. So the practical shape of December is: one workspace for thinking and working, one separate Codex view for coding, and one shared budget underneath all of it.
I read this as OpenAI tidying up a product line that grew faster than its interface. ChatGPT Work brought agentic capabilities into the main app, Codex grew out of the coding side, and the tabs multiplied. Merging them is sensible — and it points the same way as the budget change: one subscription, one workspace, one meter.
Why this hits developers hardest
For someone who only chats, this change barely registers. For someone who runs Codex hard, it changes the arithmetic of the whole day.
Agentic coding is token-hungry in a way chatting is not. A single Codex session can read dozens of files, run tests, fail, retry, and rewrite — each step consuming input and output tokens at a pace no conversation matches. Under separate budgets, that hunger was contained: a brutal afternoon of agent work left your chat allowance untouched for the evening's planning. Under a shared budget, that same afternoon eats the allowance you would have spent asking questions, drafting documentation, or reviewing a design.
Here is the scenario I am actually worried about, because it is my normal Tuesday. Morning: a long Codex session refactoring a WordPress plugin, lots of file reads and test runs. Afternoon: back in ChatGPT to plan the next feature and draft a client explanation. Today, those are two budgets. In December, the morning's agent run decides how much of the afternoon's thinking I can afford. If I hit the limit mid-afternoon, my options are waiting for a reset, upgrading, or buying credits — none of which helps a deadline that afternoon.
There is one piece of good news for the top tiers. On October 29, OpenAI doubled the Work and Codex allowances for the Pro 100, Pro 200 and Pro 500 tiers, so the shared budget lands on limits that were just raised. Plus and Business users did not get that cushion, and Pro users still face weekly caps whose interaction with the combined allowance OpenAI has not spelled out. The doubling softens the change at the top; it does not remove it anywhere.
What OpenAI has not told us yet
A responsible article should be explicit about the gaps, because the gaps are where bad planning happens. Here is what remains unpublished as of November 2:
- No per-plan token numbers. Nobody outside OpenAI knows how many tokens a Plus, Pro or Business allowance actually contains. Without that number, you cannot convert "a heavy Codex afternoon" into a percentage of your budget.
- No exact start date. "Starting in December" is the whole date. Plan for December 1, because planning for a later date and being wrong is the expensive mistake.
- No per-model burn rates. We know advanced models and complex tasks cost more, but not how much more. A GPT-6.1 session and a frontier-model session are not interchangeable, and the exchange rate is unpublished.
- No credit pricing for overflow. Buying credits is listed as an option at the limit, but the Help Center page does not say what credits cost in December, which features they unlock, or whether the subscription credits benefit announced on October 29 applies to Codex overflow at all.
- No clarity on existing windows. Codex usage today involves a 5-hour window and a weekly window on some plans, and agentic features already share an allowance across Codex, ChatGPT Work and related tools. How those windows map onto the new single token pool is not documented.
None of this makes the change a scandal. It makes it a change you should prepare for with measurements, not assumptions.
How I am preparing before December
I am not switching plans or tools on the basis of an announcement with this many blanks in it. I am doing four practical things, and they are all things you can do this week.
First, I am measuring my own burn. Starting now, I check Settings, then Usage, once a week and note roughly how my usage splits between chat and Codex. OpenAI's only reliable capacity-planning input is your own measured usage, because no published number exists. Two weeks of notes will tell me more about December than any speculation.
Second, I am moving lightweight conversations to Instant chat. Quick questions, short rewrites, and throwaway brainstorming do not need a frontier model, and Instant stays outside the meter. That is free budget, and I intend to use it.
Third, I am batching Codex work instead of dripping it through the day. One focused agent session with a clear task, a tight scope, and a review step at the end burns fewer tokens than five scattered sessions that each re-read the same codebase from scratch. This is good practice anyway; December just gives it a price tag.
Fourth, I am talking to the people I share work with about who uses what. On a shared account, one developer's heavy agent afternoon is now everyone's problem. Agreeing rough norms — heavy refactors in the morning, planning later, Instant for the small stuff — beats discovering the shared budget is empty at 4 p.m.
What I am not doing is panic-upgrading. Pro's doubled allowances are real, but upgrading to solve a budgeting problem you have not measured yet is how you pay twice: once for the bigger plan, once for the credits you still buy because your workflow never changed.
Where this fits in the bigger picture
Zoom out, and December's change is part of a longer OpenAI story about subscriptions becoming platforms. The October 29 API credits benefit for eligible ChatGPT subscribers, the doubled Pro allowances, the Work tab merger, and now a single token pool all point to one subscription that covers chatting, coding, and agentic work, metered in one currency. That is a cleaner product — and one where the meter's details still belong to OpenAI, not the user.
Conclusion
OpenAI's December change is small on paper and significant in practice: one token allowance for ChatGPT and Codex, Instant chat exempt, one dashboard to watch, and a merged Chat and Work experience rolling out alongside it. The confirmed facts are few, the missing numbers are many, and the right response is measurement, not guesswork. Check your usage weekly, move the small stuff to Instant, batch your Codex sessions, and plan for December 1. Do that, and the shared budget is a manageable constraint. Ignore it, and the first time you discover it will be the afternoon it runs out.
Frequently asked questions
When do ChatGPT and Codex usage limits merge?
Starting in December 2026, according to OpenAI's Help Center article "Understanding ChatGPT and Codex usage limits," updated October 31, 2026. No exact day has been published, so plan for December 1.
Which plans are affected?
Plus, Pro and Business plans. Their usage allowance will be shared across ChatGPT and Codex, measured in tokens.
What counts against the shared allowance?
ChatGPT conversations and Codex coding runs both draw from it. How much a request uses depends on the model and the task — advanced models and complex tasks use more.
Does Instant chat count?
No. OpenAI says "Instant chat will not draw from your usage," and its subscriber email describes Instant as unlimited on paid plans.
What happens when I hit the limit?
You can wait for the reset, upgrade your plan, or buy credits to continue using eligible features. Combined remaining usage and reset times appear in Settings, under Usage.
Is Codex disappearing into ChatGPT?
No. The Chat and Work tabs are merging into one experience, starting with a small test group, but Codex remains a separate view inside the ChatGPT desktop app.
