ChatGPT and Codex will share one usage limit from December

If you pay for ChatGPT and also use Codex, OpenAI's coding agent, you have so far been living with two separate meters. Chats in ChatGPT had their own limits, and Codex sessions had theirs. That is about to change. In a help center article that appeared on 31 October 2026, OpenAI says that starting in December, the usage allowance on its Plus, Pro and Business plans "will be shared across ChatGPT and Codex as we bring the agentic capabilities of Work into the main ChatGPT experience."

The change sounds like small print, but it touches almost everyone who pays for ChatGPT and writes code with it. It decides how much you can do in a day, which tasks eat into that budget and what happens when you hit the ceiling. Below is what OpenAI has actually said, what it means in practice and which questions are still unanswered.

The three things you need to know

1. One pool for chats and code. The core of the announcement is a single allowance. Today, a Plus or Pro subscriber who runs out of Codex usage can usually keep chatting, and the other way around, because the two products are metered separately. From December, both will draw from the same bucket. A long afternoon of Codex refactoring will leave less room for chats later that day, and a heavy research session in ChatGPT will leave less for Codex. OpenAI connects the change to a product move: the agentic capabilities of Work, its mode for longer multi step tasks, are being folded into the main ChatGPT experience. When chat, agents and coding live in one app, one shared budget is the logical next step from OpenAI's point of view.

2. Tokens instead of message counts. The second change is how usage is measured. OpenAI writes that usage for ChatGPT and Codex "will be measured in tokens, the units of information a model processes and generates." How much of the allowance a request uses "will depend 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." There is one clear exception: "Instant chat will not draw from your usage." Quick everyday questions answered by the fast Instant model stay outside the meter, while reasoning models, agent runs and coding tasks are what consume the budget.

3. The open questions. OpenAI has not said on which day in December the switch happens, and the help page does not list how large the shared allowance is for Plus, Pro or Business. What it does describe is the fallback: if you reach your limit, "you can wait for it to reset, upgrade your plan, or buy credits to continue using eligible features." Your combined remaining usage and the reset times will be visible in Settings, under Usage.

Why OpenAI is doing this

Separate limits made sense when ChatGPT was a chat app and Codex was a separate tool for developers. That line has been blurring for months. ChatGPT can now run agents that browse, write files and work through long tasks, and Codex sessions can be started from the same account and the same subscription. Keeping two independent counters for what is increasingly one product creates odd situations, for example a user who is blocked in one place while plenty of capacity sits unused in the other.

A token based allowance also matches the real cost of running these systems much better than a fixed number of messages. A one line question to a small model and a 40 minute agent run that reads an entire code base are both "one request" in a message count, but they are very different workloads on OpenAI's servers. Measuring tokens lets OpenAI charge the budget according to what a task actually costs, and it gives the company a single unit it can use across chat, agents and coding.

For OpenAI there is also a commercial logic. A shared pool with an explicit "buy credits" option turns the moment of hitting a limit into a sales opportunity rather than a dead end. Users who regularly run out can top up for specific features instead of having to jump to a more expensive plan right away.

What changes for different users

Casual ChatGPT users on Plus. If you mostly ask quick questions and use the fast model, the change may hardly be noticeable, because Instant chat will not draw from your usage at all. The budget matters once you lean on reasoning models, deep research style tasks or agent runs. OpenAI says it expects "most users won't need to wait for their usage to reset more often than they do now."

Developers who live in Codex. This group is the most affected. Codex sessions tend to be long and token heavy: the agent reads files, runs commands, writes patches and iterates. Under the new model, that work and your regular ChatGPT use will compete for the same allowance. If you were used to burning through your Codex limit and then switching to ChatGPT for planning or writing, that escape route disappears. The Settings page with combined usage will become an important place to check before starting a big job.

Teams on Business plans. For Business workspaces the change is mostly about planning. Admins will want to know how quickly typical users hit the shared ceiling and whether buying credits is cheaper than moving people to a higher plan. Until OpenAI publishes the actual numbers, that calculation has to wait.

Pro subscribers. Pro already offers the most generous limits, and nothing on the help page suggests that this changes. But Pro users are often the heaviest agent and Codex users, so the token based meter is where they will notice first if the new budget is tighter or looser than today.

How to prepare before December

There is not much you need to do right now, but a few habits will help once the shared allowance is live:

• Pick the model for the task. Since more advanced models use more of the allowance, it is worth using the fast Instant model for simple questions and saving reasoning models for problems that really need them.
• Keep Codex tasks focused. Smaller, clearly scoped tasks usually read and write fewer tokens than one huge open ended request. Splitting a big change into steps can stretch your budget further.
• Watch Settings, then Usage. OpenAI says your combined remaining usage and reset times will appear there. Checking it before a long agent run avoids being cut off halfway through.
• Know your fallback. When you hit the limit you can wait for the reset, upgrade, or buy credits for eligible features. Deciding in advance which of these you prefer saves frustration in the middle of a deadline.

What we still do not know

The help page is short, and it leaves several practical questions open. OpenAI has not published the size of the shared allowance for each plan, so nobody can yet compare the new budget with the current separate limits. It has not said how many tokens typical tasks consume, which makes it hard to estimate how many Codex sessions fit into a day. The exact start date in December is not public, and neither is the full list of "eligible features" that can be extended with credits.

It is also unclear how the change interacts with API usage, which is billed separately, and with plans that are not named on the page. The article only mentions Plus, Pro and Business. Free users and Enterprise customers are not covered by the announcement as published.

How the news spread

The help page went live quietly on 31 October 2026 and was picked up the same day on Hacker News and in tech media. TokenPost reported on the shift toward a combined, token based allowance, and other outlets followed. Some users reported receiving emails from OpenAI about bringing chat and Work together in one experience. OpenAI has not yet published a blog post or a detailed pricing table for the new system, so the help center article remains the primary source for now.

Our take

A shared, token based allowance is a reasonable way to price a product in which chat, agents and code are merging. It is fairer to people who ask quick questions, because those will not count at all, and it reflects the real cost of long agent runs. The risk sits with heavy Codex users, who lose the comfort of two separate buckets and will have to watch one meter much more closely.

Whether this is good or bad news depends entirely on numbers OpenAI has not yet shared. If the combined allowance is close to the sum of today's separate limits, most people will barely notice. If it is smaller, December will feel like a price increase disguised as a simplification. The promise that most users will not hit resets more often than now is encouraging, but it is a promise, not a figure. We will update this story when OpenAI publishes the exact date and the allowance for each plan.

Sources

• OpenAI Help Center: Understanding ChatGPT and Codex usage limits
• OpenAI Help Center: Using Codex with your ChatGPT plan
• TokenPost: report on the shared ChatGPT and Codex allowance