OpenAI to Introduce Unified Usage Limit for ChatGPT and Codex
Starting December 2026, Plus, Pro, and Business subscribers will use a single unified limit for ChatGPT and Codex. This limit applies to the web chat and the software agent but will not consume Instant quick responses.
OpenAI will abandon counting messages: instead, the tokens processed and generated by the model will be tracked. The amount deducted depends on the model and task complexity, and the remaining balance can be viewed under "Settings → Usage." Once the limit is exhausted, users will be prompted to wait for the reset, choose a more expensive plan, or purchase additional credits.
The size of the new quotas, the exact transition date in December, and the accounting rules for reasoning modes Medium, High, and Extra High are not yet known. The company explains the decision as a result of transferring ChatGPT Work’s agent functions into the main ChatGPT interface and claims that most users will not need to wait for limit restoration more frequently than before.
Previously, ChatGPT and Codex had independent limits: activity in one service did not reduce the other’s available capacity. After the merger, long conversations with resource-intensive models may reduce access to Codex, while intensive coding work could limit interactions in ChatGPT.
