ChatGPT and Codex Share One Allowance From December
Two meters become one in December. The people who lose are the ones who code heavily and chat heavily on the same account, and nobody has published how big the pool is.
From December 2026, ChatGPT Plus, Pro and Business plans will share a single usage allowance across ChatGPT and Codex, measured in tokens rather than in separate per-product limits. OpenAI's help article on ChatGPT and Codex usage limits, updated on 31 October, states the change directly and adds that Instant chat will not draw from the allowance. No exact date in December and no per-plan token amounts have been published.1
This is a change in how usage is counted rather than in how much is granted, and those two things are easy to confuse. The distinction is the whole story: OpenAI has not said the pool will be bigger or smaller, only that two meters become one.2
What the help article says
| Question | What OpenAI states |
|---|---|
| Which plans | Plus, Pro and Business |
| What is shared | One allowance covering ChatGPT and Codex |
| How it is measured | Tokens, the units a model reads and writes |
| What is exempt | Instant chat does not draw from your usage |
| Where to check | Settings, then Usage, for combined remaining usage and reset times |
| At the limit | Wait for the reset, upgrade the plan, or buy credits |
| When | December 2026, no exact date given |
Prices measured on 27 October 2026. Vendor plan terms, allowances and prices change without notice, so treat every figure on this page as a dated snapshot rather than a permanent fact. Where a number here differs from the vendor's own pricing page, the vendor's page is correct and the figure above is simply older than it.
The stated reason is consolidation: OpenAI says the change arrives "as we bring the agentic capabilities of Work into the main ChatGPT experience". Read plainly, the products are converging and the billing is following.
OpenAI also says it is "preparing carefully for this transition" and expects "most users won't need to wait for their usage to reset more often than they do now". That is a forecast, not a commitment, and it is the only statement in the article that addresses whether anyone's experience gets worse.
What is already shared, and what joins
The pooling is not new in kind. On personal plans, Codex, ChatGPT Work, Excel and PowerPoint already draw on the same agentic allowance when those features are available on your plan. December adds chat with the advanced models to that pool.
| Use | Today | From December |
|---|---|---|
| Codex | Shared agentic allowance | One shared allowance |
| ChatGPT Work, Word, Excel, PowerPoint | Shared agentic allowance | One shared allowance |
| Chat with advanced models | Separate chat limits | One shared allowance |
| Instant chat | Separate | Does not count |
| Free and Go | Not affected | Not affected |
The Instant carve-out is the detail worth holding on to. Instant is the fast, everyday mode, and it is powered by the same model free users get. In practice, OpenAI has drawn a line between the expensive reasoning work, which now shares one pool, and the cheap conversational layer, which does not.
Who this actually affects
Anyone who uses both products on the same paid account, and the effect runs in both directions.
Heavy Codex users lose the most optionality. A long agentic coding session now draws from the pool that also funds advanced chat. Before December, a weekend of Codex work left chat untouched. Afterwards it does not.
Heavy chat users on advanced models have the mirror problem. Extended reasoning sessions reduce what is available for coding work later in the month.
Teams on Business plans arguably gain. One number to forecast beats two allowances that run dry at different times, and a single pool cannot strand capacity in a bucket nobody is using that week.
Free and Go users are untouched. The article does not mention either plan, and both already run the same GPT-5.6 Luna model with volume as the only difference. If you are on the free tier, nothing here reaches you.
The number nobody has published
There is no per-plan token allowance in the article. That absence is the most important fact in this story, because it makes every "my limit is now X" claim you will read this week unsourced.
It also means the honest answer to "is this a downgrade?" is that nobody outside OpenAI can say yet. A merged pool that is sized generously could leave most users better off by removing stranded capacity. A merged pool sized to the old agentic allowance would squeeze anyone who chats on advanced models.
What can be estimated today is only the shape of the spread. OpenAI's current Work and Codex guidance gives a Plus five-hour window as roughly 5 to 45 messages with GPT-6 Astra against 350 to 3,000 with GPT-6 Luna. Those are ranges from the present system, not December figures, and they illustrate the one thing that is certain: model choice moves consumption by orders of magnitude. On a shared pool, picking the expensive model becomes a budgeting decision rather than a preference.
What to do before December
Measure now, while the meters are still separate. Note your typical weekly Codex usage and your typical advanced-chat usage in Settings and Usage. Once the pool merges, that combined figure is what you will be managing, and having a baseline means you will know within a week whether the new arrangement is tighter.
Decide which side of your usage is discretionary. If Codex work is revenue and chat is convenience, the trade-off under pressure is obvious. If the reverse, plan to drop to Instant for routine questions, since that is the one path that spends nothing from the shared pool.
And read the article again in December. OpenAI has given itself a month-long window with no date, no amounts and no confirmation that the figures will be published before the change takes effect. In a year where one announced ChatGPT cut was cancelled on the morning it was due, a commitment this vague is worth re-checking rather than planning around.
For context on what changed just before this, OpenAI cancelled the Pro 200 allowance cut on 29 October, which raised limits rather than pooling them. The two changes point in opposite directions on the same week's evidence, and only one of them comes with numbers.
Frequently asked questions
When does the shared allowance start?
December 2026. OpenAI has not published a specific date.
Which plans are affected?
Plus, Pro and Business. The article does not mention Free, Go, Enterprise or Edu.
Does normal chat use the shared pool?
Instant chat does not draw from the allowance. Chat with the advanced models does, from December.
How much allowance does each plan get?
OpenAI has not published token amounts for any plan.
What happens if I hit the limit?
You can wait for the reset, upgrade your plan, or buy credits for eligible features.
Where do I see what is left?
Settings, then Usage, which will show combined remaining usage and reset times.
Will I hit my limit more often?
OpenAI expects most users will not need to wait for resets more often than they do now. That is an expectation without published data behind it.
Scope, token measurement, the Instant chat exemption, plan coverage and the absence of published allowance sizes verified in OpenAI's "Understanding ChatGPT and Codex usage limits" help article, updated 31 October 2026 and read on 2 November 2026. The GPT-6 Astra and Luna message ranges come from OpenAI's current Work and Codex guidance and are not December figures.
Sources
- OpenAI: Understanding ChatGPT and Codex usage limits: help.openai.com/en/articles/20001625-understanding-chatgpt-and-codex-usage-limits. Used for: The shared token allowance, Instant chat exemption, plan coverage and the absence of published amounts.
- OpenAI ChatGPT release notes: help.openai.com/en/articles/6825453-chatgpt-release-notes. Used for: The 29 October reversal context and the current Work and Codex guidance.
All sources are the vendor's own plan or documentation pages, checked on 27 October 2026. Plan terms change without notice, so confirm current limits before you rely on them.
