As of about 6:45 AM IST, Monday 2 November 2026. Claims that could not be independently verified are flagged as such.
OpenAI is changing how ChatGPT usage limits work. Starting in December, Plus, Pro and Business subscribers will have a single usage allowance, measured in tokens, shared across ChatGPT and its Codex coding agent, as the company folds the agentic features of its separate "Work" mode into the main ChatGPT experience. The change appears on an updated OpenAI Help Center page and was first reported by The Decoder on 1 November.
What OpenAI says is changing
According to OpenAI's support article "Understanding ChatGPT and Codex usage limits":
- From December, the usage allowance for Plus, Pro and Business plans "will be shared across ChatGPT and Codex as we bring the agentic capabilities of Work into the main ChatGPT experience".
- Usage will be measured in tokens, the units of information a model processes and generates. More advanced models and complex tasks "generally use more of your allowance".
- "Instant chat will not draw from your usage."
- Combined remaining usage and reset times will appear in Settings, under Usage. When users reach the limit, they can wait for a reset, upgrade their plan or buy credits.
OpenAI says it is "preparing carefully" for the transition and expects that "most users won't need to wait for their usage to reset more often than they do now".
Why it matters
Until now, ChatGPT users have juggled separate modes and limits. The Decoder noted that users currently choose between two operating modes and 11 model settings. A single budget simplifies that, but it also means heavy agent tasks, coding sessions in Codex and advanced-model chats all draw on the same pool. The Decoder argued this signals the end of "all-you-can-prompt" pricing for power users, since everyday Instant chats stay unmetered while premium work is rationed by tokens.
The Decoder also reported that OpenAI recently doubled usage limits on higher-priced plans, presumably to prepare for the merger. HorizonDesk has not independently confirmed the size of that increase or the exact December start date, which OpenAI has not specified beyond "December".
A wider industry shift
The move mirrors a broader trend of AI companies converging chat and agent products. Anthropic has already combined Chat and "Cowork" in Claude, according to The Decoder. For businesses, token-based allowances make costs more predictable for the vendor but harder to forecast for teams whose usage varies widely week to week.
What users can do now
- Check how much you use advanced models, agent tasks and Codex today, since those will count most against the shared budget.
- Use Instant models for quick everyday questions, which OpenAI says will not draw from the allowance.
- Watch Settings, under Usage, once the change goes live, and note your reset times before starting long agent jobs.
Separate change: Daybreak Blue training data (unconfirmed)
The Decoder also reported, based on emails to members of an OpenAI Trusted Access programme, that OpenAI will begin using "Daybreak Blue" cybersecurity session data to train models from 29 November, with a separate opt-out promised. HorizonDesk has not seen those emails and could not independently confirm the details.
Related reading on HorizonDesk
- ChatGPT's Chrome extension and Work mode
- OpenAI models and Codex in a Google Cloud deal
- OpenAI stops live internet training for frontier models
