ChatGPT and Codex usage limits: what the December shared allowance means

From the video's published English caption track. Prices said in the video were correct on the day it was made; see the comparison above for today's check.

0:00 ChatGPT’s December usage change

If you pay twenty dollars a month for ChatGPT Plus, the way your usage gets counted is changing in December. ChatGPT and Codex are going to share one allowance, and the good news is that for a lot of people this probably won't change much day to day. But if you lean on the heavier features, it's worth understanding now, before you run into a limit in the middle of something important.

0:20 What OpenAI actually announced

So let's start with what OpenAI has actually said, because there's already a bit of confusion going around. On its help page, OpenAI says that starting in December, the usage allowance on Plus, Pro and Business plans will be shared across ChatGPT and Codex. You might have seen headlines saying December thirty-first. That date isn't on OpenAI's page.

It just says "starting in December," so if you see a specific day quoted somewhere, take it with a grain of salt. The reason this is happening is that OpenAI is bringing the agent features from ChatGPT Work into the main ChatGPT experience. Work is the mode that goes off and handles multi-step jobs for you, things like building a presentation or researching something across the web. Codex is their coding agent.

1:02 Instant chat and the shared pool

Now, here's the part I think matters most. OpenAI says Instant chat won't draw from your usage at all. So if your ChatGPT habit is mostly asking a quick question and getting a quick answer, this change shouldn't touch you much. That doesn't necessarily mean Instant is unlimited, by the way.

It just means it isn't coming out of this shared allowance. So who should pay attention? It helps to think about it a bit like a phone plan. The quick everyday stuff sits off to the side, but the heavy stuff, like the more advanced models, long tasks, and agents working away in the background, is what eats into your allowance.

OpenAI says usage will be measured in tokens, which are basically the chunks of text a model reads and writes, and that more advanced models and complex tasks generally use more of it. Let me make that real with three people.

1:46 Three everyday usage examples

The first is a business owner who uses ChatGPT to reply to customer emails and write the occasional social post. That's mostly quick chat, so honestly, this change will barely register for them. The second is a creator who uses Work to build slide decks, research sponsors and draft long scripts. They're already using up allowance today, and they'll keep doing that in December.

And the third is someone who does both, heavy agent work during the day and then coding in Codex at night. That's the person who really needs to keep an eye on this.

2:14 December allowances: what is still unknown

So how big is the bucket? This is where I want to be straight with you. OpenAI hasn't published the December numbers yet. It does publish rough estimates for how far Plus goes today in Work and Codex, and the ranges are wide, because a short question and a giant task just aren't the same size.

They're estimates, not promises, and nobody knows yet whether they'll change in December. What OpenAI has said is that it expects most users won't need to wait for their usage to reset more often than they do now. That's encouraging, but it's an expectation, not a guarantee. So what can you actually do about it?

2:44 Check your usage before upgrading

There are three things I'd do. First, get familiar with your usage page. OpenAI says that in Settings, under Usage, you'll be able to see your combined remaining usage and when it resets. Once December rolls around, glance at it every week or so.

It takes about thirty seconds, and it tells you far more than any headline will.

3:06 Match the model to the task

Second, match the tool to the job. If it's a quick question, Instant is the easy choice, since it doesn't come out of the shared allowance. If it's a big task, pick your model on purpose. OpenAI's own guidance says higher reasoning settings and Fast mode use more of your allowance, and it suggests lighter models for routine, repetitive work.

3:27 Resets, credits and upgrade options

Third, know your options before you need them. If you do hit the limit, OpenAI says you can wait for it to reset, buy credits to keep using eligible features, or upgrade your plan. Plus is twenty dollars a month, and the next step up is Pro, which starts at a hundred. That's a big jump, so I wouldn't make it because of one frustrating afternoon.

One more detail that's easy to miss.

3:46 Pro allowances versus Plus

In late October, OpenAI doubled the Work and Codex allowance on some Pro plans, and on Plus, in their words, "Plus usage is unchanged." So Plus didn't get a boost ahead of this switch. So here's how I'd think about the decision. If you mostly use quick chat, stay on Plus.

4:02 Should you stay, upgrade or switch?

This probably won't affect you much, though it's still worth a look at your usage page once December arrives. If you use Work or Codex a few times a week, staying on Plus is likely fine, but check your usage during December and see how it actually goes. If you're hitting the limit every week and it's genuinely costing you work, that's when Pro, or buying credits now and then, starts to make sense. Weigh it against what an hour of your time is worth.

And if you're on Plus but barely use any of this, it's fair to ask whether you need Plus at all. ChatGPT Go is eight dollars a month in the US. Basically, don't upgrade out of fear. Check your usage in December, see what the numbers actually say, and decide from there.

And if you'd like an update when OpenAI publishes the real allowances, subscribe, and we'll walk you through it.