ChatGPT and Codex shared usage limit: OpenAI counts tokens from December

OpenAI has announced that from December, Plus, Pro and Business plans will get a shared ChatGPT and Codex usage limit, measured in tokens. The change is described in an OpenAI Help Center article updated on October 31, and overnight from November 1 to 2 OpenAI emailed subscribers about it.

What is actually changing?

Two things at once. The first is how ChatGPT looks: the Work tab, the agent mode in which ChatGPT carries out longer tasks on its own, is moving into regular chat. According to OpenAI’s email, in the coming weeks we’ll be able to hand tasks directly to ChatGPT, from building websites and presentations to taking actions on the web, without switching to the Work tab.

ChatGPT is also meant to decide on its own when to reach for the available tools, and to break complex tasks into smaller steps. In practice, that’s an AI agent built into an ordinary conversation.

The second thing is billing. According to the Help Center, ChatGPT and Codex usage will be measured in tokens (“Your usage for ChatGPT and Codex will be measured in tokens”) and will draw from a single shared allowance.

How will the shared ChatGPT and Codex limit work?

Here is what follows from OpenAI’s email and Help Center article, among other things:

1. The limit applies to the Plus, Pro and Business plans and is set to start in December. OpenAI hasn’t given an exact date.
2. Usage depends on the model and the task: more advanced models and complex tasks use up a larger share of the limit.
3. Chatting with Instant models doesn’t count toward the limit. The email says the plan includes unlimited Instant use.
4. We can check how much is left and when the limit resets under Settings → Usage.
5. When the limit runs out, we can wait for it to reset, upgrade to a higher plan, or buy credits to keep using eligible features.

Interestingly, OpenAI reassures users that most of them won’t have to wait for a reset more often than they do today (“We expect most users won’t need to wait for their usage to reset more often than they do now”).

Why is this making waves?

Because it’s the end of ChatGPT as an “all-you-can-eat” subscription. Until now, chat and Codex had separate limits. From now on, every conversation with a stronger model, every agent task and every Codex session will draw from the same pool of tokens.

The Decoder calls this the end of “all-you-can-prompt” pricing and points out that thanks to free Instant, ChatGPT remains a flat-rate subscription “on paper”. According to the editors, though, heavy use of the cheaper plans will more often end in weaker answers.

For developers, this is a big change: anyone who spends hours in Codex today and then chats at length in ChatGPT will have to split one allowance between both.

What should we watch out for?

First, the size of the limit. There aren’t many details yet: OpenAI hasn’t said how many tokens each plan will get, or how the current chat and Codex limits will translate into the shared pool.

Second, the models we use every day. Since Instant is unlimited and stronger models use up more, it makes sense to save them for tasks that really need them (check in Settings how much a typical workday uses up before the change lands in December).

Third, credits. Being able to buy more after hitting the limit is convenient, but it’s also a back door to bills higher than the subscription itself. On the other hand, we get an agent in regular chat with no mode switching, which for many people may be more convenient than a separate Work tab.

We know similar dilemmas from Claude’s limits: Anthropic recently dropped the 5-hour limit on the Claude Ultra plan, and measurements of token usage in Claude Code showed what agents spend their allowance on.

FAQ

When does the shared ChatGPT and Codex limit start?

According to the OpenAI Help Center, in December 2026. OpenAI hasn’t given an exact date, and the email to subscribers talks about “the coming weeks”.

Will ChatGPT Plus have a token limit?

Yes. Plus, Pro and Business will get a shared allowance for ChatGPT and Codex, measured in tokens. OpenAI hasn’t yet published the size of the limit for each plan.

Do Instant models have a limit?

No. Chatting with Instant models doesn’t count toward the shared limit.

What can I do when I run out of my ChatGPT limit?

Wait for it to reset, upgrade to a higher plan, or buy credits. The Settings → Usage section shows your remaining limit and when it resets.