OpenAI has announced that starting in December, subscribers on ChatGPT Plus, Pro and Business plans will share a single usage allowance across ChatGPT and Codex. The change lands alongside a bigger interface shift: the agentic Work experience is being folded into ChatGPT's main interface, ending the split between a chat tab and a work tab that the company introduced only months ago.
The mechanics are straightforward once you read the fine print. Usage in both ChatGPT and Codex will be metered in tokens, and how fast the pool drains depends on which model you pick and how complex the task is — a long autonomous coding run will burn through far more than a handful of questions. Instant chat is exempt and does not count against the shared allowance. Users will see their combined remaining balance and the next reset time in Settings, and when the pool runs dry the options are the familiar trio: wait for the automatic reset, upgrade the plan, or buy extra credits.
The interface consolidation is already visible to some users. OpenAI says it will roll out a unified "Chat and Work" experience to a small group in the coming weeks before expanding, letting people pull files, connectors and browser-running tasks into the same conversation where they chat. Codex will keep a standalone entry point in the ChatGPT desktop app. President Greg Brockman has admitted the current split app felt messy, and has pledged that there will be no separate Work tab by the end of the year.
The timing is no accident. Last week OpenAI roughly doubled Work and Codex allowances across its Pro tiers at unchanged prices and walked back a planned reduction, after heavy users revolted. It has also quietly started enforcing device limits on paid accounts — a hard cap on how many personal subscriptions can be logged in on one machine across Go, Plus and the Pro 100, 200 and 500 tiers, with Business, Enterprise, Edu and free accounts exempt. Logging out of one account does not free a slot for another, a rule clearly aimed at teams stacking multiple subscriptions to pool their quotas.
Put the three moves together and the strategy reads clearly: one interface, one metering system, one pool, and no easy loopholes. Subscription economics at OpenAI are being rebuilt around agentic usage rather than message counts. That is rational at a moment when Anthropic — fresh off filing for an IPO — is competing for the same enterprise buyers, and when every doubled allowance has to be paid for somewhere in compute costs.
For most subscribers the merged pool should be painless; OpenAI's own estimate is that most people will not wait for resets any more often than they do today. The friction will concentrate at the top: developers who run Codex for hours a day will now watch coding work and casual chat draw from the same budget, and any future price or quota adjustment will be visible in one place. The company that turned token limits into a public relations problem last week has decided the cleanest fix is to make all its products share the meter.
Comments (0)
Log in to join the discussion
Log InNo comments yet