OpenAI rewrote its usage-limits help article on Sunday, and buried in it is a change that hits anyone running Codex alongside ChatGPT: from December, the two share a single allowance. Plus, Pro and Business are all affected.
OpenAI’s stated reason is that the agentic capabilities of Work are moving into the main ChatGPT experience. When one product absorbs the other, two budgets become one.
The counting unit changes to tokens
That is the second half of the change, and it matters more than the merge itself. Instead of messages or tasks, OpenAI will count tokens, the units a model reads and writes in. What a job costs now depends on the model you pick and how heavy the work is. Stronger models and longer runs eat more.
One carve-out is spelled out: instant chat will not draw from the allowance. Your combined remaining usage and reset times are supposed to show up under Settings → Usage. Once you hit the limit, you can wait for the reset, change plan, or buy credits.
OpenAI says it is preparing the transition carefully and expects most users won’t wait for a reset more often than they do today. That is an expectation, not a promise, and it is exactly the sentence people will hold the company to in December.
Two vendors, two directions
Anthropic did the opposite on Friday: the Max plans are now Max and Ultra, and the 5-hour limit is gone for verified accounts. I wrote about that yesterday. One vendor pulls a gate out, the other pours two allowances into one pot and starts counting tokens.
In practice: if you use Codex for long agent runs and ChatGPT on the side for thinking, from December you have to decide where the budget goes. Token accounting makes consumption more honest and much harder to estimate. With a message cap you know where you stand. With tokens you find out after an agent has spent an hour reading files.
The number is the thing the help article doesn’t give. How big the shared allowance is per plan, OpenAI doesn’t say, and until that lands the merge is neither good news nor bad news. It’s a change with a hole in the middle of it.