What is OpenAI changing in December 2026?
OpenAI will combine the usage allowances for ChatGPT and Codex on its main paid plans. The change begins in December, according to a help-center notice that OpenAI updated over the weekend. The notice covers Plus, Pro and Business subscribers.
The notice is short and direct. It says: "Starting in December, Plus, Pro, and Business plans' usage allowance will be shared across ChatGPT and Codex." OpenAI frames the change as part of bringing the agentic capabilities of Work into the main ChatGPT experience. Work is the company's mode for longer research and deliverable tasks.
Under the new setup, usage for both products is measured in tokens. Tokens are the units of information a model reads and writes. A prompt, an attached file, chat history, tool results and the model's reply all consume them. OpenAI says advanced models and complex tasks generally use more of the allowance than simple ones.
Token metering is not the same as counting messages. A short question with no attachments may use a few hundred tokens. A long document, a large codebase or a multi-step tool call can use thousands. That is why OpenAI's guidance focuses on the model and the task rather than a single message quota.
One carve-out matters for casual users. OpenAI states that Instant chat will not draw from the shared usage. The company also says subscribers can see combined remaining usage and reset times in Settings, under Usage. Users who reach the limit have three options: wait for a reset, upgrade their plan, or buy credits for eligible features.
Which plans share the allowance, and what do they cost?
The notice names Plus, Pro and Business. Those plans sit at very different price points, so the same pooling rule will feel different depending on what a subscriber pays. Current prices are listed on OpenAI's pricing page and in its Codex pricing documentation.
Plus costs $20 a month and includes expanded Codex usage. Pro is now sold in three tiers at $100, $200 and $500 per month. The $500 tier is the only one that includes Ultrafast mode, according to OpenAI's Pro plan help page. Pro 200 is also available to new subscribers again.
The Pro tiers also differ in included usage. OpenAI says Pro 200 includes more usage than Pro 100, and Pro 500 offers the highest included usage of the three. Subscribers choosing a tier should compare the Work and Codex allowance, not just the headline price.
Business is priced per seat. OpenAI lists $20 per user per month when billed annually, or $25 per user per month on monthly billing. Business bundles ChatGPT, Work and Codex, and includes admin controls such as SAML single sign-on and multi-factor authentication.
Plus also unlocks more Codex surfaces than the cheaper plans. OpenAI's Codex pricing documentation lists Codex on the web, in the command line, in the IDE extension and on iOS, plus cloud integrations such as automatic code review and Slack. Those features matter when a task runs in the background rather than in a chat window.
Free and Go are not named in the December notice. That does not mean Codex disappears from cheaper plans. OpenAI's Codex help page says Codex is included across ChatGPT plans, including Free and Go. Codex Cloud, which runs tasks on OpenAI-managed computers, is limited to eligible Plus, Pro, Business, Enterprise, Healthcare and Education accounts.
- •Free: $0, limited Codex access
- •Go: $8 per month, more messages and uploads
- •Plus: $20 per month, expanded Codex usage
- •Pro: $100, $200 or $500 per month
- •Business: $20 per user per month annually, or $25 monthly
How many messages will one shared token allowance buy?
There is no single answer, because token use depends on the model, the task and the amount of context. OpenAI's documentation gives ranges instead of fixed message counts. The estimates below apply to local messages in a five-hour window for Plus and Standard Business seats.
The spread is wide. On GPT-6 Astra, the most expensive model in the list, the estimate is 5 to 45 local messages per five-hour period. GPT-6.1 Sol is listed at 15 to 160, and GPT-6.1 Terra at 35 to 285. GPT-6 Sol is listed at 15 to 150, while the cheaper GPT-6 Luna is listed at 350 to 3,000.
Those figures are not guarantees. OpenAI explicitly says the numbers are estimates rather than fixed message limits and that users should check their usage dashboard for current limits and reset times. Cloud tasks can consume more of the allowance than local messages, and weekly limits may also apply.
Where a task runs also matters. OpenAI says usage depends on the model, where the task runs, its complexity, context, reasoning effort, speed and tools. A long-running task can use substantially more than a short request. That makes a coding agent's background work harder to predict than a single chat reply.
Cloud work adds a second dimension. OpenAI says cloud tasks can continue while a user's computer is asleep. That convenience also means a task can keep consuming allowance after the person who started it has moved on to something else.
Pro plans currently have no five-hour limit, according to the same documentation. That does not make them unlimited. OpenAI says any unlimited access included with Pro is subject to abuse guardrails. Codex usage has run on five-hour and weekly windows, and a full banked reset refreshes both windows while changing the weekly reset date.
- •GPT-6 Astra: 5 to 45 local messages per five hours
- •GPT-6.1 Sol: 15 to 160 local messages per five hours
- •GPT-6.1 Terra: 35 to 285 local messages per five hours
- •GPT-6 Sol: 15 to 150 local messages per five hours
- •GPT-6 Luna: 350 to 3,000 local messages per five hours
Why is OpenAI pooling the two allowances?
OpenAI's stated reason is product consolidation. It wants the agentic features of Work inside the main ChatGPT experience, and a single allowance is the billing model that matches that design. The help-center notice does not offer a detailed cost rationale beyond that.
Some sharing already exists. OpenAI's pricing documentation says Work usage inside ChatGPT uses the same pricing, credits and limits as Codex. It also says usage limits are shared with other agentic features once pricing for those features takes effect, starting with ChatGPT for Excel on Plus and Pro. OpenAI's release notes say ChatGPT for Word draws from the shared usage allowance for Codex and other premium features.
The company has also been adjusting the size of those pools. On 29 October, OpenAI doubled the Work and Codex allowance included in Pro 100 and Pro 500 plans. It canceled a Pro 200 reduction that had been announced in a September 29 email, and doubled the lower Pro 200 allowance to match the higher one. The same help page says Plus usage was unchanged.
That timing matters. The doubling raised the baseline for some Pro subscribers just before the pools are merged. It also shows that OpenAI can change allowance sizes without changing monthly prices, which is one reason the December numbers are hard to predict.
The product merger has been moving for months. OpenAI's July desktop update added a way to switch between Chat and Work, then combined both conversation types in a unified Recents list. The December change completes that consolidation on the billing side.
The Decoder, an AI news site that covered the December notice, argues that pooling will likely lower quality for heavy users on cheaper plans. It writes that the change is "likely to result in a decline in ChatGPT's quality for heavy users on lower-cost plans." That is analysis rather than an OpenAI statement. The outcome depends on allowance sizes that OpenAI has not published.
What happens when the shared allowance runs out?
OpenAI lists three routes: wait for the allowance to reset, upgrade to a higher plan, or buy credits. The Codex help page adds that eligible Plus and Pro users can buy credits without changing plans. Credit prices depend on the plan or agreement, according to OpenAI's pricing documentation.
Some activity stays outside the pool. OpenAI says auto-review safety checks are free and do not count toward plan usage limits. ChatGPT file-upload, image-generation and voice limits have separate caps, reset periods and banners, and they do not apply to Codex. Starting a new Codex chat does not reset usage limits.
Credits add another layer. OpenAI's release notes describe auto top-up for eligible Plus and Pro users, which refills a shared credit balance when it falls below a chosen minimum. The feature is managed from Codex Settings. Subscribers who enable it should treat it as a spending control, not a way to avoid the allowance.
Subscribers should also watch their own habits. A developer who runs long cloud tasks will consume allowance differently from someone who asks short questions. OpenAI advises users to open the usage dashboard when they approach a limit. It shows which allowance is exhausted, any credit balance and the next reset time.
The practical effect of pooling is that unused capacity no longer sits idle in a separate bucket. A subscriber who barely uses Codex but chats often may benefit from the flexibility. A subscriber who pushes both products hard may reach the combined ceiling sooner, unless OpenAI raises the total allowance when December arrives.
What is still unconfirmed about the December change?
Several important details are missing. OpenAI has not given an exact start date beyond December. It has not published the size of the shared allowance for each plan, and it has not said whether the total allowance will rise, fall or stay the same when the pools are merged.
The treatment of Free, Go, Enterprise and Edu is also unclear. The notice names Plus, Pro and Business only. OpenAI's pricing documentation says Enterprise and Edu plans without flexible pricing have the same per-seat limits as Plus for most features. Enterprise customers with flexible pricing have no fixed rate limits and pay through credits.
OpenAI's confidence about disruption is untested. The notice says: "We expect most users won't need to wait for their usage to reset more often than they do now." That is a prediction, not a published measurement. Subscribers will only be able to judge it after the change goes live.
The price of extra credits is another gap. OpenAI's pricing documentation says credit purchase prices and applicable discounts depend on the plan or agreement. Subscribers therefore cannot calculate the cost of exceeding the shared allowance until OpenAI publishes the rates for each plan.
Some users are already uneasy. A thread on OpenAI's developer community, opened on 31 October, asks the company to keep the two limits separate. One Plus subscriber wrote that extending a shared token allowance to advanced chat could interrupt workflows. The thread is a record of concern, not evidence about how the final system will behave.
What should ChatGPT and Codex users watch next?
The first thing to watch is the December rollout notice. OpenAI will need to publish allowance sizes, credit prices and the exact date before subscribers can judge the change. Those details should appear in the same help-center article and in the ChatGPT release notes.
The second is the usage dashboard. Once combined usage appears in Settings, subscribers can compare their real consumption with the published estimates. A Plus user who mostly sends short prompts may see little difference. A user who relies on long Codex sessions and advanced models has more reason to track the numbers.
Users can also build a baseline before December. A week of normal work gives a rough sense of tokens per task. That personal baseline is more useful than a vendor's range when deciding whether Plus, Pro 100 or Business fits a workload.
The third is plan economics. Pooling makes the Pro tiers easier to compare on one axis, because chat and coding draw from the same budget. It also makes credits more important as a pressure valve. Buyers should check whether extra credits cost less than moving up a tier for their workload.
Finally, watch how OpenAI describes the change after launch. If the company publishes token budgets and reset behavior, the December shift will be easy to evaluate. If it does not, subscribers will be left comparing anecdotes. For now, the verified facts are simple: one allowance, three paid plans, a December start and no published size for the pool.