OpenAI is changing how paid subscribers use ChatGPT and Codex. Starting in December 2026, Plus, Pro and Business plans will draw from a shared usage allowance across the two products, according to an updated OpenAI Help Center notice.
The important exception is Instant chat, which OpenAI says will not consume that allowance. For people who combine demanding conversations with coding work, the change makes understanding usage more important. Here is what the new ChatGPT and Codex shared usage limits mean, what remains unknown, and how to prepare without making unnecessary subscription changes.
Background: some OpenAI tools already share usage
This is not the first shared allowance in OpenAI’s products. Its Codex documentation says ChatGPT Work and Codex already share usage, pricing and credits. Other supported agentic features can also draw from overlapping allowances once their pricing takes effect.
The December announcement extends the shared arrangement across ChatGPT and Codex as OpenAI brings Work’s agentic capabilities into the main ChatGPT experience. The distinction matters: an existing shared budget for delegated work is not the same as adding more conversational activity to that budget.
The New Stack reported on the change on 2 November, highlighting concerns from subscribers who use ChatGPT for planning and research alongside Codex for software development.
What changes in December?
Plus, Pro and Business are explicitly named
OpenAI’s notice names these three plan families. It does not establish an identical December change for Free, Go, Enterprise or Edu. Users on other plans should check their own product notices rather than assume every account will follow the same rules.
The notice says the transition starts in December but gives no precise day. It also does not publish the new allowance sizes for each named plan.
Usage will be measured in tokens
Tokens are units of information a model processes and generates. OpenAI says consumption depends on the model selected and the task requested; more advanced models and complex jobs generally use more allowance.
That means counting prompts is not a dependable way to forecast capacity. Codex’s existing documentation explains that context, reasoning, tools and task complexity can affect usage. A brief request involving substantial background material may consume more than its length suggests.
Instant chat stays outside the shared allowance
OpenAI explicitly exempts Instant chat from this usage pool. That makes it relevant for everyday questions that do not need a more demanding workflow. However, exemption from this particular allowance should not be read as a promise that every Instant feature has no other limits.
Subscribers will be able to view combined remaining usage and reset times under Settings → Usage. On reaching a limit, the stated options are waiting for a reset, upgrading, or purchasing credits for eligible features.
Why the shared allowance matters
The practical issue is competition between activities. Under a common pool, substantial coding work can leave less capacity for other metered ChatGPT tasks, while demanding conversational work can reduce what remains for Codex.
Consider a developer who researches a migration in ChatGPT, delegates implementation to Codex, and then asks for a detailed review. Those activities may look like separate tools, but their allowance becomes connected. The effect depends on the final allocation and the work performed—not simply how many products are open.
OpenAI says it expects most users will not need to wait for resets more frequently than today. That is a stated expectation, not a published guarantee of identical capacity for every workflow. Without the new allocations, claims of either a universal downgrade or a universal improvement are premature.
Practical steps for users and businesses
Record your current usage patterns
Before the transition, note which tasks matter most and when you approach existing limits. After the change reaches your account, compare similar working days using the usage display. This gives you a better basis for decisions than comparing unrelated sessions.
Match the tool and model to the task
Use Instant chat where its capabilities are sufficient, and reserve demanding workflows for tasks that genuinely benefit from them. For coding, supply relevant files and clear acceptance criteria rather than an unnecessarily broad brief.
OpenAI’s Codex guidance recommends limiting irrelevant context and matching output to the required result. These are sensible efficiency measures, not guarantees of a fixed number of extra tasks.
Review team spending before buying credits
Businesses should identify who can approve extra usage and how staff should handle a limit during important work. The notice names Business plans, but it does not justify assuming every organisation has identical billing controls or an organisation-wide pooled allowance.
Australian teams should check the actual currency, taxes and terms displayed for their account before approving additional spending. There is no need to invent an Australian-dollar estimate for allowance sizes that have not been disclosed.
Risks, limitations and unanswered questions
The largest gap is transparency. OpenAI has not stated the new token allocations in this notice, so users cannot reliably calculate their future capacity. A visible remaining balance will help, but product-by-product and task-level detail would make unexpected consumption easier to explain.
Token measurement also does not automatically mean subscriptions are becoming unrestricted pay-as-you-go API accounts. Included subscription usage, purchased credits and API-key billing are distinct arrangements. OpenAI’s Codex pricing documentation explicitly warns against using API token prices to estimate included subscription tasks.
Another unknown is how existing reset structures will map to the December arrangement. Do not assume current limits, promotional allowances or reset rules remain unchanged unless updated documentation confirms it.
What to watch next
Watch for a precise rollout date, plan-specific allocations and guidance explaining how existing limits translate into the combined system. Clear reporting of which activities consume the pool would also help subscribers manage their work.
Check the official notice and your account’s usage page before changing plans. Documentation and account-specific messages are more dependable than unofficial calculators based on undisclosed conversion rules.
Conclusion
ChatGPT and Codex shared usage limits will connect more paid AI activity from December, while leaving Instant chat outside the pool. The sensible response is to monitor real consumption, scope demanding tasks carefully and wait for the missing plan details before deciding whether extra capacity is necessary.
Sources
- OpenAI Help Center — Understanding ChatGPT and Codex usage limits
- OpenAI developer documentation — Codex pricing and usage limits
- The New Stack — OpenAI’s new shared usage limit pits ChatGPT against Codex