OpenAI will combine ChatGPT and Codex usage into one allowance for Plus, Pro and Business subscribers starting in December, as the company moves ChatGPT Work’s agentic features into the main Chat interface.
The change means demanding work carried out in ChatGPT and coding sessions run through Codex will draw from the same pool. Ordinary Instant chat will remain outside that allowance, according to new OpenAI guidance. Subscribers will be able to see their combined remaining usage and reset times under Settings > Usage.
OpenAI has not given an exact December launch date or published a universal message count. It says usage will be measured in tokens, with the amount consumed varying by model, task complexity and the work required. That makes this a consolidation of metering as well as a product-interface change: a long coding job in Codex and a multi-step task started in ChatGPT will no longer live behind separate usage counters.
What changes in December
OpenAI’s current Work experience is designed for jobs that can use files, connected tools and multiple steps. In an email reviewed by BleepingComputer, the company described a combined Chat and Work test that will let selected users move from a conversation into execution without changing tabs. Examples included building a website, creating a presentation and taking actions on the web while the user reviews progress.
The test is expected to begin with a small group before expanding. Codex will still exist as a separate experience in the ChatGPT desktop app, but its eligible usage will feed the same allowance described in OpenAI’s updated help article.
| Activity | December treatment |
|---|---|
| Instant chat | Does not draw from the shared allowance |
| Agentic tasks in ChatGPT | Draw from the combined ChatGPT and Codex pool |
| Codex coding work | Draws from the same combined pool |
| OpenAI API use | Billed separately at API rates |
The first three rows are covered by OpenAI’s new usage guidance. Separate API billing is stated in the company’s current Work and Codex documentation; promotional API credits, where offered, are also separate from ChatGPT usage credits.
Why the meter may move at different speeds
A shared pool does not turn every prompt into a fixed charge. OpenAI says larger inputs and outputs, more capable models, higher reasoning settings, Fast mode and multi-step tasks can all consume more allowance. Cloud tasks may also use more than short local exchanges.
That distinction matters once Work-style tasks appear inside a familiar chat window. Uploading a large set of files, asking an agent to research across connected services, or sending it through a long coding workflow can use substantially more capacity than asking a brief question. The interface may look continuous even when the work behind it is not.
OpenAI’s existing guidance recommends checking the usage page before a large task and selecting the model and reasoning level before starting. It also warns that switching models does not restore allowance inside a shared pool. The upcoming meter should make the cross-product effect visible, but the company has not yet published a complete December rate table showing how every ChatGPT mode will translate into usage.
The Pro increase is a separate change
The December consolidation follows an October 29 increase in included Work and Codex capacity for higher-priced plans. OpenAI’s release notes say Pro 100 and Pro 500 now receive twice their previous Work and Codex allowance. The company also canceled a planned reduction for Pro 200 subscribers; customers on the lower Pro 200 allowance received a doubling that brought them up to the retained level.
Those increases should not be read as separate ChatGPT and Codex buckets. They raise the relevant included capacity now, while the December change determines how eligible ChatGPT and Codex activity will share it. OpenAI has not disclosed exact numerical limits for every tier in the new combined setup, so users should rely on the meter attached to their own account rather than estimates circulating elsewhere.
What subscribers should check before rollout
Plus, Pro and Business users who depend on Codex or Work for long sessions should record their normal usage pattern before December. The practical baseline is visible under Settings > Usage: remaining allowance, any five-hour window, the weekly reset time and available credit options.
- Keep routine questions in Instant when it meets the need, because OpenAI says Instant chat will not consume the shared pool.
- Choose the model for the task. A cheaper, faster model can preserve allowance for work that genuinely needs a frontier model or higher reasoning effort.
- Check the meter before long jobs. A Codex session can affect how much capacity remains for an agentic task started later in ChatGPT.
- Separate included usage from credits. Credits pay for supported work after included limits are exhausted; they do not increase model access.
- Read reset terms before buying. OpenAI says an eligible paid instant reset restores the applicable allowance immediately and moves the weekly reset schedule forward. It is not an extra banked week.
Paid instant resets are currently offered only to eligible personal Plus and Pro accounts, with availability varying by account and billing country. Business, Enterprise and Edu accounts are excluded from that specific reset option.
The most important unanswered details are the exact rollout day, the final per-plan allowances and how individual non-Instant ChatGPT modes will be labeled in the usage view. Until OpenAI publishes those rates, the confirmed change is narrower but consequential: from December, eligible agentic ChatGPT work and Codex will compete for the same included capacity, while Instant chat remains outside it.
