OpenAI has confirmed a significant change to how its subscriptions are metered: starting in December, Plus, Pro, and Business plans will share a single usage allowance across ChatGPT and Codex, the company's coding agent. The change is laid out in OpenAI's updated help-center documentation, "Understanding ChatGPT and Codex usage limits," and coincides with OpenAI folding the agentic capabilities of its Work product into the main ChatGPT experience.
Until now, heavy users have juggled two separate buckets: one for chatting and agent work in ChatGPT, and another for Codex, the agent developers use for coding tasks. From December, both draw from one pool — a simplification on paper, but one that changes the math for anyone who uses both services intensively. For more context on this story, see our ongoing AI industry coverage.
What Changes in December
OpenAI's documentation is direct about the mechanics. "Starting in December, Plus, Pro, and Business plans' usage allowance will be shared across ChatGPT and Codex as we bring the agentic capabilities of Work into the main ChatGPT experience," the help page states.
Under the new system, usage across both products is measured in tokens — the units of information a model processes and generates. How quickly the allowance drains depends on two factors: the model chosen and the task asked of it. More advanced models and more complex tasks consume more of the allowance, so a long Codex session on a frontier model will bite into the same budget that powers ChatGPT conversations.
There is one important carve-out: instant chat will not draw from the usage allowance at all. Light, quick interactions remain effectively unmetered, which blunts — though does not eliminate — the impact on casual users.
The token-based approach is a shift in framing as much as mechanics. Subscription plans have historically been marketed with message counts and hourly caps, numbers that were easy to reason about but crude as a measure of actual compute. Tokens map directly to what a model reads and writes, which lets OpenAI price a two-hour autonomous coding session and a set of complex research queries on the same scale — and lets subscribers see a single remaining-balance figure instead of per-product counters.
What Subscribers Will See
The company says it is preparing carefully for the transition to prevent disruption to existing workflows, and that it expects most users will not need to wait for their usage to reset more often than they do today. That framing matters: OpenAI's last several usage-limit changes were received poorly by its most engaged customers.
Once the change takes effect, subscribers will see their combined remaining usage and reset times in Settings, under Usage. Hitting the cap presents three options, per the documentation: wait for the allowance to reset, upgrade the plan, or buy credits to continue using eligible features.
Why OpenAI Is Unifying the Pools
The stated reason is product consolidation. OpenAI has been pulling its agentic capabilities — including the Work-oriented features that let ChatGPT act autonomously on tasks — into the main ChatGPT app rather than distributing them across separate products. Codex, meanwhile, has grown from a code-completion tool into a full autonomous coding agent. With both surfaces converging on the same underlying agent infrastructure, maintaining two independent usage pools made less sense.
The unified allowance also fits OpenAI's broader push to make ChatGPT the single entry point for knowledge work. The company recently introduced account-level changes that let subscribers buy credits for use beyond subscription limits — the same mechanism the December change leans on when users exhaust their pooled allowance.
A Sensitive Topic for Power Users
Usage limits are the most closely watched dimension of any ChatGPT subscription, and OpenAI has history here. Hacker News has carried a running thread of complaints about allowance changes, including a Pro-plan reduction earlier this month and discussion of the new combined limits over the weekend. Power users — particularly developers running Codex for hours at a time — are the group most likely to feel the merge, since their Codex consumption will now compete directly with their ChatGPT usage.
For OpenAI, the consolidation simplifies communication and probably smooths capacity planning across products. For subscribers, the practical advice is to audit how usage splits between ChatGPT and Codex today: anyone whose combined consumption already pushes against one of the two existing caps may hit the merged ceiling faster, while users who lean on one product at a time may see little difference.
There are a few concrete steps worth taking before December. Check current consumption in Settings, under Usage, to establish a baseline for each product. Note which models the heaviest workflows use, since advanced models drain the shared pool fastest — a task that runs comfortably on a mid-tier model today may be a poor trade-off once it competes with the rest of a subscriber's allowance. And teams on Business plans should identify which automations actually need frontier models and which can run cheaper, because the combined metering makes that distinction financially visible for the first time.
The credits mechanism also deserves attention. OpenAI's documentation confirms that users who reach their limit can buy credits to continue using eligible features rather than waiting for a reset — an option that effectively turns the shared allowance into a base tier with metered overage. For individual developers, that may be the difference between a blocked afternoon and an uninterrupted deployment window; for finance teams, it introduces a variable cost line that fixed subscriptions previously avoided.
The change takes effect in December across Plus, Pro, and Business plans. Team and Enterprise arrangements, which are typically contractually metered, are not addressed in the current documentation.
Source: OpenAI Help Center, updated October 31, 2026.---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
