Here’s a change worth knowing about if you pay for ChatGPT. From December, your chats and your Codex coding work will draw from one shared pool, not two separate ones. Instant chat is the main exception — it stays outside the pool. OpenAI has not yet said which day in December this starts, or how big each plan’s pool will be.
What happened
On 31 October 2026, OpenAI quietly updated its Help Center page on ChatGPT and Codex usage limits. Reports based on that page started to appear on 1 November. The message is short and clear: starting in December, Plus, Pro, and Business plans get one shared allowance across ChatGPT and Codex.
Why now? OpenAI says it is bringing the work features of Codex into the main ChatGPT experience. So the two products are joining up, and their usage limits are joining up too.
Now, what does “measured in tokens” actually mean for you? A token is just a small piece of text. Your message counts, and so do files you attach, earlier messages in the chat, tool results, and the answer itself. Bigger models chew through more tokens. Longer tasks do too. A long agent job can easily cost far more than a quick question — which is exactly why heavy users should pay attention here.
OpenAI’s own take is calm. It expects most users will not need to wait for a reset more often than they do today. If you hammer both chat and Codex in the same week, you are the person most likely to notice this change.
What counts and what does not count
OpenAI has spelled out some rules, and left other bits open.
What counts toward the shared pool:
- Chat in ChatGPT with any model other than Instant.
- Codex work, whether you use the desktop app, the command line tool, the editor extension, the web app, or Codex Cloud.
- ChatGPT Work, ChatGPT for Excel, and Workspace Agents. These already share the Codex allowance today, so nothing new there.
What does not count:
- Instant chat. OpenAI puts it plainly: “Instant chat will not draw from your usage.”
- Codex with your own API key. That is billed at API rates instead, so it sits outside your plan pool.
- Codex auto-review safety checks, when you sign in with ChatGPT.
Still not stated:
- The exact December day.
- The size of the pool for each plan.
- How image generation, file uploads, and regular Chat Voice in ChatGPT will fit in. Right now those have their own separate limits.
- How Free and Go plans are affected. The reports name Plus, Pro, and Business only.
- How weekly caps on Pro plans will work with the shared pool.
In ChatGPT, open Settings, then Usage. In Codex, open Usage and Billing. After the change, that page will show your combined remaining usage and reset times.
Why this matters for freelancers and students
Picture a normal freelance Thursday. Morning goes to client emails and proposals in ChatGPT, afternoon goes to fixing a client’s code in Codex — and suddenly, come Friday, the chat side feels tight. Both jobs drank from the same pot all week.
Students, same story, different desk. Say you lean on ChatGPT to plan an essay in the morning, then switch to Codex in the evening to learn code or build a tiny project. Under one pool, those two habits now pull against each other. Short study chats on their own? Barely a ripple. But stack a long coding run on top of long chats with a big model, and — well, you can guess what happens to the pool.
Then comes the money side, which is worth a minute of your time. Once you hit the limit, three doors open: wait for the reset, move to a bigger plan, or buy credits. Credits sound handy, but they expire after 12 months, you usually cannot get a refund, and you cannot hand them to another account. Your plan pool always gets spent first, credits only step in afterwards, so buy a little first, and decide after a real December week.
For context, on 29 October 2026, OpenAI doubled the Work and Codex allowance on Pro 100 and Pro 500, and dropped a planned cut for Pro 200. You can read our earlier post on Pro limits for the full story on that change, which came just before this shared-pool news.
How to make your allowance last
Start with Instant, the quiet hero here. Because Instant chat sits outside the shared pool, it is perfect for the little stuff: a quick fact, a short rewrite, a simple translation. Big models? Keep those for work that genuinely needs the muscle. Every small question you move to Instant is pool left over for Codex later.
Coding jobs are the next lever. Codex lets you pick your own model, so don’t reach for the biggest one out of habit. Light models give you far more messages in a five-hour window, and for a routine job that is usually plenty. Honestly, the biggest model is like hiring a crane to hang a picture — impressive, but overkill most days.
Speed modes deserve a look too. Fast burns the pool at 2.5 times the Standard rate, Ultrafast at 8 times, and only Pro $500 users even get Ultrafast. Unless a deadline is breathing down your neck, Standard does the same job for less. Waiting a few extra seconds is a cheap trade.
Small habits that cut waste
Long chats are sneaky. Each new message quietly resends everything before it, so the longer a chat runs, the pricier every reply gets. My fix: when a job ends, close the chat. Open a fresh one for the next job, with a one-line note on what is already done. No, this does not refill your limit — it just stops the slow leak.
Agents need the same discipline, in writing. A vague brief — “clean up this project” — can wander for ages, touching files you never meant it to touch. Instead, spell it out: these files only, this exact end result, and stop to check with me before anything big. Narrow briefs, cheaper runs.
And images? In Codex they are hungry. An image turn eats the pool about 3 to 5 times faster than a similar turn without one, and OpenAI still has not said how ChatGPT image limits will fit the new pool. Until December brings clarity, treat images as a treat, not a habit.
What to do when you hit the limit
First, check the reset time. Open Settings, then Usage, in ChatGPT, or Usage and Billing in Codex. And do not ask Support for a reset — OpenAI says Support does not reset ChatGPT or Codex usage limits, full stop.
Second, if the task can wait, waiting costs nothing.
Third, if you really must keep working, you can buy credits from Settings, then Usage, in ChatGPT, or from Usage and Billing in Codex. Just remember two things first: credits expire after 12 months, and a negative balance from an earlier task can eat a new purchase before you get any use out of it. If you turn on auto-reload, set a maximum monthly spend — or leave it off if you want tight control.
Fourth, check if you hold a saved reset from a referral offer. Some Plus and Pro users get one. Using it refreshes both the five-hour window and the weekly window for Codex, but it also moves your weekly reset date, so look in your profile menu for this offer before you spend money.
What is still not clear
Three facts are still missing, and they are big ones. OpenAI has not named the exact December day, it has not published the pool size for each plan, and it has not said how some ChatGPT limits, like image and voice limits, will fit into the shared pool.
An OpenAI engineer wrote on 28 October 2026 that the team is bringing Chat and Work together with care, and will share more soon. Until OpenAI publishes the numbers, treat any exact quota you see online as unconfirmed. The safe facts are the ones in this article: one shared pool, Plus, Pro, and Business plans, a December start, and Instant chat excluded.
If you use GPT-6.1 Terra for routine computer work, you may also want our earlier post on GPT-6.1 Terra — Terra is one of the lighter, lower-cost options that can stretch a shared pool further.
The bottom line
From December, paid ChatGPT users get one usage pool for chat and Codex, not two separate pools, and Instant chat stays outside it. Most light users will likely feel little change. But if you mix long chats with long Codex tasks, start checking Settings, then Usage, each week now. That habit gives you real numbers.