News · 2026-11-01

OpenAI says ChatGPT and Codex will share a usage allowance starting in December

OpenAI says ChatGPT Plus, Pro and Business subscribers will share a token-based usage allowance across ChatGPT and Codex starting in December. Instant chat will not draw from that allowance, but OpenAI has not published the pool sizes or a specific December start date. The change makes coding and other advanced ChatGPT work compete within the same stated allowance.

Key facts

• The announcement names three plans: Plus, Pro and Business.
• The shared allowance starts in December 2026; no exact day or token totals are specified.
• OpenAI says, “Instant chat will not draw from your usage.”
• The primary source is OpenAI’s usage-limits Help Center article.

For subscribers who use both products, the important change is the relationship between their activities. A long coding session and an advanced chat task will draw from a common allowance instead of remaining separate usage categories. The source does not provide enough numerical detail to say whether a particular subscriber will gain or lose practical capacity.

A shared mobile-data plan is a useful analogy. A laptop and phone can both consume one household allowance, so heavy use on one device changes what remains for the other. Here the activities are coding and chat, and the accounting unit is tokens rather than bytes. The analogy explains the coupling; it does not imply a one-for-one price between tasks or models.

OpenAI says the amount used depends on the model and task, with more advanced models and complex work generally consuming more. That makes a simple message count misleading. A short request can initiate a long sequence of reading, reasoning and tool use. Our inference-cost lesson explains why the visible length of a prompt does not measure all the work behind it.

The Instant exception is clear but narrow. OpenAI’s quoted sentence establishes exclusion from this allowance. It does not use the word unlimited or rule out other controls. A supplied Reddit headline converted that exception into an unlimited-use claim; the primary article does not support that stronger reading.

OpenAI says subscribers will be able to see combined remaining usage and reset times under Settings and Usage. It lists waiting for a reset, upgrading, or purchasing credits for eligible features as options when the allowance is exhausted. Those are the company’s stated plans and instructions, not observations from a completed rollout.

The company also expects most people will not have to wait for resets more often than they do now. That is a prospective expectation. Without the plan totals, conversion rates and real post-transition usage patterns, it cannot be independently checked today. A user running many background coding tasks may have a different workload from someone mostly using ordinary chat.

The broader product context is ChatGPT Work. OpenAI’s July 9 product announcement describes Work as an agent in ChatGPT that acts across apps and files and stays with multistep projects. The December article says those agentic capabilities are coming into the main ChatGPT experience.

That does not establish every future interface decision. It does not say whether Work will remain a distinct selectable experience or precisely how the integration will appear. Describing the change as a merger of Chat and Work can be convenient shorthand, but the concrete usage policy is the shared ChatGPT-and-Codex allowance for the named plans.

A different October story is easy to blend into this one. OpenAI’s October 29 release note revised Work and Codex allowances and reversed a previously announced reduction for Pro 200. BleepingComputer’s October 29 report covers that reversal. It does not document the new December pool, and combining the two can create a false impression that a doubling applies uniformly to the forthcoming arrangement.

The November 1 Developer Community notice repeats the shared-allowance timing. In the Hacker News discussion, readers question the absence of published budgets and the loss of separation between coding and chat. Those comments identify real unanswered questions, but they are not representative polling or proof of a capacity cut.

The caveat is therefore practical rather than rhetorical: the structure is announced, while its numerical consequences are not. Subscribers can identify which activities become linked, but cannot yet calculate how many demanding sessions their plan will support. The useful accountability checkpoint is December’s rollout and the publication or display of an allowance that can be compared with actual work.

Primary source, verified: read the paper →

Ground Truth. (2026, November 1). OpenAI says ChatGPT and Codex will share a usage allowance starting in December. Ground Truth. https://groundtruth.day/news/openai-chatgpt-codex-shared-allowance-december.html

@misc{groundtruth:openai-chatgpt-codex-shared-allowance-december,
  title  = {OpenAI says ChatGPT and Codex will share a usage allowance starting in December},
  author = {{Ground Truth}},
  year   = {2026},
  month  = {nov},
  url    = {https://groundtruth.day/news/openai-chatgpt-codex-shared-allowance-december.html}
}

Topics: openai · chatgpt · codex · subscriptions · product-policy