OpenAI Just Turned Your ChatGPT Subscription Into a Meter. December Is When the Dial Starts.

The most consequential thing OpenAI did this week was not a model launch. It was a support page edit.

“Starting in December, Plus, Pro, and Business plans’ usage allowance will be shared across ChatGPT and Codex as we bring the agentic capabilities of Work into the main ChatGPT experience.” That sentence, sitting on OpenAI’s help center under “Understanding ChatGPT and Codex usage limits,” is the end of an era most people didn’t realize they were in.

Two separate buckets become one. Your coding agent and your chat assistant now drink from the same glass. And the glass is measured in tokens, which means the flat monthly fee you have been treating as an all-you-can-eat plan has just become a metered utility with a cap you cannot see.

What Actually Changes in December

Read the support page carefully, because the details determine who gets hurt and who doesn’t.

Usage for ChatGPT and Codex is now measured in tokens — the units of information a model processes and generates. How much of your allowance you spend depends on two things: the model you pick and the task you assign it. More advanced models and complex tasks generally use more. That is OpenAI’s own phrasing, and it is a polite way of saying the good stuff costs more of your month.

The escape hatch is Instant chat, which “will not draw from your usage.” On paper this keeps ChatGPT a flat-rate product. In practice it creates a two-tier reality: the fast, cheaper model stays unmetered, and everything you actually wanted to do — the reasoning, the agentic work, the long tasks that made you pay for Pro instead of staying free — runs on the meter.

And the transition math is the part OpenAI is least specific about. “We expect most users won’t need to wait for their usage to reset more often than they do now.” That is not a promise. That is an expectation, published in advance, so that when your reset cycle shortens in January there is a document saying they thought it probably wouldn’t.

From December, Plus, Pro and Business subscribers get one token allowance covering both ChatGPT and Codex — a heavy agentic session in either tool now shortens what’s left for the other.

The Timeline Nobody Announced

This was not a launch event. It came out in pieces, which is worth noting because the pieces tell you how it was sequenced.

OpenAI emailed ChatGPT subscribers about the change, and BleepingComputer obtained that email. TikTok-style test group first: a small set of users gets the merged Chat and Work experience in the coming weeks, moving from casual conversation into multi-step tasks without switching modes. “You can bring in your files and plugins and ask ChatGPT to build a website, create a presentation, or do things for you on the web, while reviewing and guiding its work.”

Then the allowance merge. Then, separately and most tellingly, the doubling of Work and Codex allowances for Pro 100, Pro 200, and Pro 500 — effective October 29.

Look at that order again. OpenAI doubled the allowances in late October, then announced the merge in early November. That is not an accident of scheduling. You double the pool before you make everyone share it, because the alternative is a subscriber revolt in the first week.

Codex also remains a separate experience in the ChatGPT desktop app, which is a small concession to the people who do their real work in the terminal and do not want it folded into a browser tab.

The Numbers You Can Actually Compare

Here is where the pricing structures diverge in ways that matter more than the announcement itself.

OpenAI doubled Work and Codex allowances for Pro 100, Pro 200 and Pro 500 on October 29 — five weeks before the merge. The doubled pool was the preparation, not a gift.

Why OpenAI Is Doing This, Stated Plainly

Two reasons, and only one of them is about money.

The first is complexity. Users currently navigate two operating modes and eleven model settings, distinctions that OpenAI’s own critics — and, in a roundabout way, OpenAI itself — have conceded are too much. Folding Work into Chat reduces the choices to a single surface. Anthropic already did the equivalent by merging Claude Chat and Cowork into one product. When your main competitor simplifies, you simplify or you look like the complicated option.

The second is unit economics. Token-based metering converts a fixed-revenue subscription into something closer to usage-based revenue, where the heaviest users stop being a cost center and start being a sales opportunity. Credits exist for exactly this reason. Nobody buys credits on an unlimited plan.

The honesty is refreshing in a bleak sort of way. The support page says the quiet part: more advanced models and complex tasks use more of your allowance. There is no attempt to hide that the meter is measuring exactly what you paid for.

What the Community Is Saying

The reaction on X has been split along a predictable line. One user, VraserX, argued that heavy Codex usage should not eat into a ChatGPT allowance and proposed keeping separate limits with an optional shared pool. A community account that tracks Codex news, Codex Update, asked OpenAI to publish the actual token allowance for each plan before December so people could check the math for themselves.

That second request is the one OpenAI has not answered. There is a support page confirming the merge, an email confirming the merge, and no published number for what you get. Which means the first month of the new system will be a nationwide experiment in discovering your own cap by hitting it.

For anyone running a workflow on paid AI tools, that is the practical takeaway: you cannot budget for December using November’s data, because November’s data is denominated in a unit the system no longer uses.

What This Means for Your Video Workflow, Specifically

If you are on this site, you are probably not doing much Codex work. But the merge still reaches you, and here is how.

ChatGPT has quietly become the front door for a lot of video workflow: writing scripts, generating shot lists, drafting VO, building storyboards, and increasingly orchestrating the agentic tasks that stitch those outputs together. Those are exactly the tasks the support page describes as consuming more allowance — complex, multi-step, model-heavy. Your script generation was never the problem. Your script generation that also researches the topic, drafts three variants, and reformats them for each platform is now a measurable withdrawal from a shared pool.

The video generation itself sits on other subscriptions — which, if anything, makes the diversification argument stronger. The tools that merged into one meter are the ones you can no longer forecast. The ones you keep separate are the ones you can.

The practical play: draft cheap, polish expensive, batch the repetitive tasks, and watch the usage pool. A pooled allowance punishes exactly the habits that were free under separate ones.

The Honest Counter-Argument

It is only fair to note that OpenAI may be telling the truth about disruption, and that the case for the merge is not purely extractive.

The doubling of Pro allowances on October 29 was real and it came before the merge. If OpenAI wanted a pure cash grab, the sequence would have been reversed: merge first, then sell the fix. Preparing the pool first suggests at least some genuine concern about breaking heavy users’ workflows, and the “most users won’t need to wait more often than they do now” line may well hold for the majority — the median ChatGPT user is not running agentic pipelines all afternoon.

There is also a legitimate simplicity argument that gets dismissed too quickly. Two modes and eleven model settings is a genuine UX failure. If a single unified interface makes the product usable for the people who bounce off it today, the merge is a net win for them, and they are the majority.

But “most users” is doing enormous work in that sentence, and the people who pay the most are rarely in the majority. The merge is neutral for the casual user, mildly positive for the confused user, and a new expense for the power user. That is the shape of almost every platform pricing change, and it is the shape of this one.

What to Do Before December

Three moves, all of them cheap.

First, go to Settings → Usage now, before the merge, and write down what you currently consume in a normal week. You cannot compare against a baseline you never recorded, and after December the old separate counters are gone.

Second, audit your model choices. The single largest lever on your December bill is not your Codex usage — it is how often you reach for a frontier model to do something a fast model handles fine. The support page is explicit that the model choice drives consumption. That is a budget line you control.

Third, decide now whether you would rather upgrade or buy credits, because that decision will be made under pressure at the worst possible moment if you leave it. Upgrading is a predictable monthly cost. Credits are a variable expense that arrives as a surprise and always at the wrong time of the month.

And one thing to watch: whether OpenAI publishes per-plan token allowances before December, as the Codex community has asked. If they do, the merge is manageable and you can plan against it. If they don’t, everyone finds out their limit the same way — by hitting it mid-task on a Tuesday.

The Bigger Shift

Step back from the specifics and the pattern is hard to miss. Flat-rate AI is dying across the industry, and it is dying from both ends.

At the bottom, unlimited plans get quietly costed down with cheaper models and slower queues. At the top, the heavy users who extract the most value get moved onto meters. Every provider is converging on the same conclusion: token-based pricing is the only structure that scales when the product’s cost is a per-token input and the customer’s usage is unbounded.

Anthropic got there by merging products. OpenAI is getting there by merging allowances. The mechanism differs, the destination does not.

For a creator, the practical consequence is that the era of treating a single AI subscription as infinite capacity is over, and the ones who plan for it now will be the ones who don’t get their workflow cut in half by a settings screen in January. The meter is being installed. The only question left is whether you know your number before it matters.

Frequently Asked Questions

When does the shared ChatGPT and Codex allowance start?

OpenAI says the shared allowance begins in December for Plus, Pro, and Business plans, according to its support page “Understanding ChatGPT and Codex usage limits.”

Does Instant chat count toward the shared allowance?

No. OpenAI states that Instant chat “will not draw from your usage,” so the fast model remains effectively unlimited. More advanced models and complex tasks do consume the shared allowance.

Will I lose access if I hit the limit?

You will not lose the subscription. OpenAI says you can wait for the allowance to reset, upgrade your plan, or buy credits to continue using eligible features.

Where do I see my remaining usage?

Settings → Usage shows your combined remaining usage and reset times across both ChatGPT and Codex once the merge takes effect.

Why is OpenAI merging these allowances?

OpenAI is folding the agentic capabilities of Work into the main ChatGPT experience, which removes a separate mode and its separate limit. The company had also doubled Work and Codex allowances for Pro 100, Pro 200, and Pro 500 on October 29, ahead of the change.

Does Codex stay a separate app?

Yes. Codex remains a separate experience in the ChatGPT desktop app even after the allowance is shared, according to BleepingComputer.