ChatGPT and Codex Share One Usage Limit From December, Counted in Tokens

OpenAI has quietly rewritten the deal on its $20 plan. In a help-centre article updated on November 1, 2026, it says that starting in December, the usage allowance on Plus, Pro and Business "will be shared across ChatGPT and Codex", and that your usage "will be measured in tokens". If you pay for ChatGPT and have never opened Codex, this is still about you: the pool is shared in both directions.

Key Takeaways

• Three plans are named: Plus, Pro and Business. OpenAI's article does not mention Free or Go at all.
• Your limit stops counting messages and starts counting tokens. OpenAI's words: "More advanced models and complex tasks generally use more of your allowance."
• Quick questions are exempt. "Instant chat will not draw from your usage." Instant is the lowest setting on the Thinking slider, not a different product.
• The reason is that ChatGPT Work is being folded into the main app. OpenAI says the change comes "as we bring the agentic capabilities of Work into the main ChatGPT experience".
• No start date, no numbers, no conversion rate. "Starting in December" is the whole schedule. OpenAI publishes no allowance size for any plan and no tokens-per-message figure.
• OpenAI's own reassurance is a forecast, not a guarantee: it "expects most users won't need to wait for their usage to reset more often than they do now."
• One screen gets more useful: Settings then Usage will show your combined remaining usage and reset times.
• It was not announced where you would look. It is not on the ChatGPT release notes, whose newest entry on November 4, 2026 is still October 30.

Who Is Affected, and What To Do Now

What Actually Changes

Today your chatting and your agent work come out of two different buckets. From December, on Plus, Pro and Business, they come out of one. OpenAI states the current separation plainly on its models article, read live on November 4, 2026: "Usage and credit rules in Work and Codex are separate from Chat." The merge is not arriving from nowhere, either. OpenAI's credits article already says that "For Plus and Pro, Codex, ChatGPT Work, Excel, and PowerPoint can share the same agentic usage allowance", and that Word draws on it too. So four or five features already share one pool. December adds the thing you actually use every day: chat.

That is the whole shape of it. One number instead of two, and the number now includes your chats.

Why Tokens Change How It Feels

A message limit is predictable and a token limit is not. Tokens are the chunks of text a model reads and writes, so a token allowance counts the size of the work rather than the number of requests. Paste a long report and ask one question, and you have spent far more of your month than ten short questions would cost. OpenAI says exactly this: "How much of your allowance you use will depend on the model you choose and the task you ask it to do. More advanced models and complex tasks generally use more of your allowance."

OpenAI has not published a conversion, and the only token prices it does publish are for credits, on the same credits article: in ChatGPT for Word, Excel and PowerPoint, GPT-5.6 Luna costs 5 credits per million input tokens and 30 per million output, while GPT-6 Astra costs 250 and 1,250. That is a 50x spread between the cheapest and most expensive model on the same list. It is not the Plus chat allowance and should not be read as it, but it is the clearest published signal that your model choice is about to matter to your limit far more than your message count does.

The One Word Most People Will Misread

"Instant chat will not draw from your usage" is the sentence to hold on to, and "Instant" is not a figure of speech. On ChatGPT's Thinking slider, Instant is a named setting -- the lowest one. OpenAI's models article defines the levels in order: "Instant -- Fast responses for everyday questions", then Medium ("Standard thinking"), High ("Extended thinking"), Extra High and Pro. So the exemption is not "chatting is free" and it is not "short questions are free". It is a specific position on a control you may never have touched, and if you leave the slider higher than Instant, your everyday questions are inside the allowance.

That reading matters most on Plus, because Plus is the plan with the least room. OpenAI's own plan table says Plus gets Medium and High thinking and does not get Extra High or Pro, while Pro and Business get all three. So a Plus subscriber who wants more than an instant answer has exactly two settings above the exempt one, and both will be counted.

What OpenAI Has Not Said

Three gaps, and each one is the thing a reader most wants. We read the article in full on November 4, 2026 and word-diffed it against the earliest archived copy, from November 2, to confirm nothing had been edited out: the only change in two days is the "last updated" stamp.

• No date. "Starting in December" with no day, no staged rollout, no country order.
• No size. Not one allowance figure for Plus, Pro or Business, before or after.
• No conversion. Nothing that lets you translate your current experience into tokens.

OpenAI does say what happens when you hit the wall: "you can wait for it to reset, upgrade your plan, or buy credits to continue using eligible features." Credits are the pay-as-you-go route, bought from Settings then Usage, and OpenAI notes they are separate from your ChatGPT wallet balance, so gift-card funds are not usage.

Should This Change What You Pay?

Not yet, and probably not at all if you are on Free or Go. Neither plan is named, and the case for the free tier is unchanged: unlimited everyday text chats, with the caps sitting on uploads, images, voice and data analysis. If you are weighing the $20, this is one more reason to run the free plan honestly for a week first -- see is ChatGPT Plus worth it in 2026.

If you already pay for Plus, the one useful thing you can do this month costs nothing. Open Settings then Usage and look at it. When the combined number appears in December you will have something to compare it against, which is more than OpenAI is offering. And if you have never moved the Thinking slider, now is the time to learn where it is, because from December it decides what counts.