OpenAI is about to make you choose between chatting and coding. Starting in December 2026, ChatGPT and Codex will pull from a single usage pool for paying subscribers, and the company has not yet said how big that pool is.
OpenAI announced that the shared allowance will cover Plus, Pro, and Business plans. Instant chats will not count against it. Free, Go, Enterprise, and Edu accounts were not mentioned in the December change, so those users should not assume anything yet.
Usage will be measured in tokens, which means consumption depends on the model you pick and how hard the task is. A developer who runs code reviews in Codex and then asks ChatGPT for a deep technical analysis will be spending from the same balance. Heavier models and longer tasks burn through it faster. That is how API billing has always worked, but subscription users have mostly been shielded from thinking this way.
Once the policy is live, subscribers will see their combined remaining allowance and reset times under Settings → Usage. Hit the cap and you can wait for a reset, upgrade, or buy credits for eligible features. OpenAI says most people should not wait for resets more often than they do now. That claim is untested, and I would treat it as a hope rather than a promise.
The gaps are the real story. OpenAI has not published numerical limits, reset schedules, or a full explanation of how each advanced ChatGPT feature will be counted. For a team paying for both ChatGPT and Codex seats, that makes budgeting guesswork. You cannot compare the new setup to your current plan when one side of the equation is blank.
The shared pool fits a clear pattern. OpenAI is folding ChatGPT Work, its agent for longer multistep tasks, into the main ChatGPT experience, while Codex handles software development. Some subscriptions already share allowances between Work and Codex. Putting everything under one meter is the logical next step, and it mirrors where the whole market is heading: one assistant, many jobs, one bill.
There is also a competitive angle. Teams weighing Codex against Anthropic’s Claude Code now have another variable. Anthropic also ties usage limits to subscription tiers, so neither side offers unlimited use. But a unified pool can be better or worse depending on your habits. Heavy coders may feel squeezed if chat-based research eats into their agent time. Light users may barely notice.
Pricing tiers add more confusion. A separate October 29, 2026 update doubled Work and Codex allowances for Pro 100 and Pro 500 subscribers, and OpenAI canceled a planned cut for Pro 200 users. Those changes apply to current allowances, not the December arrangement. Anyone deciding between Pro 200 and Pro 500 will want new figures before paying more.
So does this matter? Yes, mostly for teams. Individuals on Plus will probably shrug until they hit a wall. Organizations running both tools should start measuring their current token-heavy workflows now, so they can check the numbers against OpenAI’s published limits the moment they appear. Until then, the policy is a direction, not a price list.
