OpenAI Merges ChatGPT Chat and Work Modes Under Token-Based Quota

What happened

OpenAI is merging its separate Chat and Work modes into a single ChatGPT interface. From December, Plus, Pro and Business plans will share one token-based quota that covers both ChatGPT and Codex.

Under the new structure, the allowance drawn down by a request depends on the model it uses. More complex models consume more of the quota, while simple instant chats do not count against it. The Decoder, which first reported the change, frames it as the end of all-you-can-prompt pricing.

The same reporting suggests that heavy users on lower-cost plans are likely to notice a drop in ChatGPT's quality, because their allowance will run down faster when they rely on more capable models.

Why it matters

This change shifts ChatGPT from a flat-rate tool toward a metered service, where the cost of each interaction reflects the computing it requires. For organisations adopting AI, that makes model choice a financial and operational decision as well as a technical one. Teams will need to understand which tasks justify a heavier model and which can be handled by a lighter, cheaper one.

The move also has clear experience implications. A shared quota across chat and work use cases could simplify the product for users who currently move between modes. However, it introduces a new source of uncertainty: people need to know how much allowance remains and why a particular task has used more of it. Predictability and transparency will shape whether users feel the service is fair.

By the numbers

• December: the month the shared, token-based quota is due to start for Plus, Pro and Business plans.
• Three subscription tiers (Plus, Pro and Business) are covered by the common quota.
• Two separate modes, Chat and Work, are being combined into one interface.

The Renascence take

Most coverage will focus on the price change, but the more important shift is in how usage is made visible. When people can see that a complex request costs more than a simple one, they start to ration the tool rather than explore it, and that quietly changes the experience.

The real design question is not whether AI usage should be metered, but whether users can predict and understand what a task will cost before they start it. Operators who roll out these plans should map their most common workflows, identify which ones genuinely need the most capable model, and set clear defaults so that employees are not left guessing. Treat quota literacy as part of the service, not a footnote in the pricing page.