OpenAI Pools ChatGPT and Codex Usage Into One Token Budget From December

Starting in December, OpenAI will share the usage allowance of its Plus, Pro, and Business plans across ChatGPT and Codex, according to its official help article on usage limits. Until now, chat and coding have sat under the same plan without drawing from a single shared allowance. The change makes that allowance one budget for both surfaces.

The allowance will be measured in tokens, the chunks of text a model reads and writes (often a word fragment or a short word). OpenAI's documentation states that how much of the allowance an interaction consumes depends on the model selected and the task requested. A short question to a lightweight model and a multi-step coding task on a larger one will therefore draw down the same budget at very different rates. Instant chat will not draw from the allowance at all.

Subscribers can view combined remaining usage and reset times under Settings → Usage. When a user reaches the limit, the documentation describes three options: wait for the allowance to reset, upgrade to a higher plan, or buy credits to keep using eligible features.

The change also touches OpenAI's agentic products. The company is bringing the agentic capabilities of Work into the main ChatGPT experience. A separate help article on ChatGPT Work and Codex notes that longer conversations and tasks in both products count toward usage limits. In other words, the metering applies to multi-turn agentic sessions, not only to individual prompts.

The pricing ladder has also been changing. OpenAI is adding a $500 per month ChatGPT Pro tier, which includes its highest usage limits and access to its Ultrafast mode, as The Verge reported on September 29. In early October, the company reopened its $200 per month tier with new usage limits, as covered in its DevDay coverage. Earlier, OpenAI's $100 per month Pro tier was reported to offer five times the Codex usage of the $20 per month Plus plan, per an April report from The Verge.

In my view, the pooled allowance is less a pricing story than a measurement story. Once chat, agentic work, and code generation share one token-denominated budget, the number on the Usage screen becomes the only meaningful constraint. Heavy Codex users and heavy chat users who previously had separate headroom will now draw on the same balance, and the reset timer becomes a planning input rather than a footnote. Whether that is a fair trade depends largely on how often a given user hits the ceiling.

The broader context here is that token-based metering is becoming the common currency of AI subscriptions. It lets a provider price a short completion and a long agentic run on the same scale, and it means model choice now affects a subscriber's bill as much as it affects the provider's inference cost. For teams on Business plans, the practical question is whether the visibility in Settings → Usage is granular enough to allocate budget across developers and workflows.

The upside, in my view, is clearer. A single visible meter with published reset times gives users a concrete way to plan long sessions, and the option to buy credits offers a path forward that does not require a plan change. The mechanics are still new. The December switch will show how users actually distribute their allowance once chat and Codex compete for it.