ChatGPT and Codex will share one token-based usage limit from December
OpenAI will combine the usage allowances of ChatGPT and Codex into a single pool starting in December, according to an updated article in the company's Help Center titled "Understanding ChatGPT and Codex usage limits." The change covers the Plus, Pro and Business plans. From that point, activity in ChatGPT and in Codex will be measured in tokens and drawn from the same remaining balance, so a subscriber who spends a long session on agentic work in ChatGPT will have less left for coding tasks in Codex, and the reverse.
The most important exception concerns ordinary conversation. OpenAI says chats with Instant models will not draw from the shared allowance, which keeps everyday question-and-answer use outside the metered pool. Everything else is weighted by model and task: the company states that more advanced models and more complex tasks consume the allowance faster, and that the amount used by a single request depends on both. OpenAI has not published a specific start date within December, and it has not published new allowance sizes for any plan.
The update arrives two days after OpenAI doubled the Work and Codex limits on its Pro plans, a move we covered in our report on the Pro and Business Premium limit increase. The company has also said that the separate Chat and Work tabs in ChatGPT will be merged into one interface, first for a small test group and then for more users, while Codex remains a distinct experience inside the ChatGPT desktop app.
··········
HOW THE SHARED CHATGPT AND CODEX ALLOWANCE WILL WORK.
The December model extends a structure that already exists for part of the product. OpenAI's current documentation states that "Work and Codex share the usage allowance included with your plan," and the official Codex pricing page adds that ChatGPT Work usage inside ChatGPT "uses the same pricing, credits, and usage limits as Codex." What changes is the scope of the pool: the Help Center article now describes one allowance spanning ChatGPT and Codex as a whole, with tokens as the unit of measurement in place of the approximate message counts that the documentation uses today.
Users will be able to see their combined remaining usage and the next reset times under Settings, in the Usage section. When the allowance runs out, OpenAI lists three options: wait for the reset, upgrade to a higher plan, or buy credits to keep using eligible features. The company says it expects most users will not have to wait for resets more often than they do now, which is a projection about typical behavior and not a guaranteed allowance.
The exclusion of Instant chats is the detail that decides how most subscribers will experience the change. Light users who mainly ask questions, draft text or summarize documents with Instant models should see no new constraint. Subscribers who mix reasoning models, agentic tasks in ChatGPT and coding work in Codex will be the ones drawing from a single balance, and for them the order in which they spend it during a five-hour or weekly window becomes a practical planning question.
··········
CHATGPT AND CODEX LIMITS BEFORE AND AFTER THE DECEMBER CHANGE.
The table below compares the rules described in OpenAI's current documentation with the rules the company has announced for December. Items OpenAI has not specified are marked as such.
........
Item | Current rules | From December |
What shares one pool | Work and Codex | ChatGPT and Codex |
Unit of measurement | Approximate messages per window | Tokens |
Instant chats | Not part of the Work and Codex pool | Do not draw from the shared allowance |
Plans covered | Plus, Pro, Business | Plus, Pro, Business |
Five-hour window | Applies to Plus and Business; none on Pro | Not specified |
Weekly limits | "May also apply" | Not specified |
At the limit | Wait, upgrade or buy credits | Wait, upgrade or buy credits |
Where to check usage | Usage dashboard | Settings, Usage |
........
The current message estimates give a sense of how unevenly the models consume an allowance. OpenAI's Codex pricing page lists, for Plus and Standard Business, local messages per five-hour period by model. The company stresses that these "are not fixed message limits" and that the usage dashboard shows the actual limits and reset times.
........
Model | Plus and Standard Business, local messages per five hours |
GPT-6 Astra | 5-45 |
GPT-6 Sol | 15-150 |
GPT-6.1 Sol | 15-160 |
GPT-6.1 Terra | 35-285 |
GPT-6 Luna | 350-3,000 |
........
At the bottom of the ranges, GPT-6 Luna allows 70 times as many messages as GPT-6 Astra (350 against 5), and at the top roughly 67 times as many (3,000 against 45). Faster speed modes widen the gap further: on included subscription usage, OpenAI applies a 2.5x multiplier for Fast mode and an 8x multiplier for Ultrafast on GPT-6 Astra and GPT-6.1 Sol. OpenAI has not said whether these estimates will be restated in token terms once the shared pool takes effect.
··········
[[ADPLUS_300x250_1]]
WHAT CHANGES FOR CHATGPT PLUS, PRO AND BUSINESS SUBSCRIBERS.
The change applies to three plan families, whose current list prices are published on OpenAI's official pricing page. Plans not named in the Help Center update keep their existing arrangements as far as OpenAI has disclosed.
........
Plan | Price (USD) | Shared pool from December |
Free | $0 | Not named |
Go | $8 a month | Not named |
Plus | $20 a month | Yes |
Pro | $100, $200 or $500 a month | Yes |
Business | $20 per user a month billed annually, $25 billed monthly | Yes |
Enterprise and Edu (flexible pricing) | By agreement, no fixed rate limits | Not named |
........
For a Plus subscriber, the five-hour and weekly windows already shape heavy Codex use, and the December pool adds agentic ChatGPT activity to the same balance. A practical way to protect coding time is to keep everyday chat on Instant models, which stay outside the pool, and to reserve the most expensive models, such as GPT-6 Astra, for tasks that need them. Users who often hit the ceiling can compare the cost of credits with the $100 Pro tier; OpenAI says credit prices and discounts depend on the plan or agreement and does not publish a single per-credit rate on the pricing page.
For Pro subscribers the starting position is more favorable. OpenAI currently applies no five-hour limit to Pro plans in Work and Codex, and their Work and Codex limits were doubled at the end of October. What OpenAI has not stated is whether the five-hour exemption carries over once ChatGPT usage joins the pool. Business administrators should expect the same mechanics for their members' seats; OpenAI's pricing page shows Standard Business with the same per-model ranges as Plus.
Readers who want the full plan-by-plan picture of message caps, tools and fair-use rules can consult our guide to ChatGPT usage limits by plan, and developers weighing Codex against competing coding agents will find the subscription costs compared in our AI coding agent pricing comparison.
··········
OPENAI'S CHAT AND WORK MERGER AND THE MOVE TO METERED USAGE.
The shared pool is tied to a product change. OpenAI says it is bringing the agentic capabilities of Work into the main ChatGPT experience, so that a user can move from a normal conversation into a multi-step task, such as building a website, preparing slides or completing an online task with their own files and plugins, without switching modes. Today ChatGPT presents separate Chat and Work tabs and a long list of model and effort settings; a single interface with a single allowance removes the need to choose a mode before starting.
The billing logic follows from that design. Once agentic tasks can start from any conversation, a separate allowance for "chat" and another for "work" would be hard to explain, and token-based metering lets OpenAI charge heavy agentic sessions according to the compute they consume while leaving Instant conversations effectively unmetered for subscribers. The same pattern is visible across the market: Anthropic removed the five-hour limit for verified accounts on its top consumer tier last week while keeping weekly limits, as described in our report on the Claude Ultra plan, and both companies now steer their heaviest users toward higher tiers or purchased capacity.
··········
CHATGPT AND CODEX SHARED LIMIT: CONFIRMED FACTS AND OPEN QUESTIONS.
OpenAI has documented the core of the change in its Help Center: from December, Plus, Pro and Business plans will draw ChatGPT and Codex activity from one allowance measured in tokens; Instant chats will not count against it; consumption will vary with the model and the complexity of the task; remaining usage and reset times will appear under Settings; and users at the limit can wait, upgrade or buy credits. The company also expects most users not to wait for resets more often than today.
Several points remain open. OpenAI has not given a start date within December, has not published allowance sizes in tokens for any plan, and has not said whether the five-hour window, the weekly window or the Pro exemption from the five-hour limit will change. It has not stated whether the per-model message estimates on its pricing page will be replaced, nor whether Free, Go, Enterprise or Edu users will move to the same model later. Until those numbers appear on the official pages, the safest reading for subscribers is that ordinary Instant chat is unaffected, while anyone who mixes reasoning models, agentic tasks and Codex on the same plan should watch the Usage page closely in the first weeks of December.
··········
FOLLOW US FOR MORE.
·····
DATA STUDIOS
·····
datastudios.org
