ChatGPT and Codex Will Share One Usage Limit in December: What Changes and What to Do
Starting in December 2026, ChatGPT chat with advanced models and Codex will draw on one shared, token-based usage allowance on the Plus, Pro and Business plans, according to an OpenAI help-center article updated on October 31. Instant chat is the stated exception and does not draw on that allowance. If you code with Codex and also chat heavily in ChatGPT, the practical effect is simple: a long coding session now leaves less room for advanced chat, and the other way round. OpenAI has not yet published the exact switch-over day or how many tokens each plan includes.
Last checked: November 3, 2026. We publish AIWITH.CHAT. OpenAI's help page refused our automated request, so the OpenAI details below come from consistent coverage of that page, which we link. Check OpenAI's article before making a purchase decision. The Gemini details were checked against Google's official changelog.
What OpenAI says is changing
- One pool. Chat with advanced models joins the allowance that Codex and other agentic features on Plus and Pro already share. Usage is measured in tokens rather than counted as messages.
- Instant chat is outside the pool. Quick answers in Instant mode do not consume the shared allowance. That is not the same as a promise of unlimited Instant use.
- Cost varies by work. Bigger models, longer tasks and more reasoning consume more of the allowance than a short question.
- Visibility. Remaining allowance and reset times appear under Settings → Usage.
- At the limit. You can wait for the reset, upgrade, or buy credits for eligible features.
- OpenAI's expectation. The coverage reports that OpenAI expects most users will not wait for resets more often than they do now. That is an expectation, not a guarantee.
What is still unknown
- The exact date in December when your account switches.
- How many tokens each plan gets, so you cannot yet compare today's limits with the new ones.
- How the 5-hour and weekly windows will apply to the combined pool.
- Whether Free, Go, Enterprise or Edu plans change in the same way.
Until those numbers appear, avoid buying an upgrade purely on the basis of the announcement. Watch Settings → Usage in your first December week and measure how fast a normal day drains it.
Who should care most
Heavy Codex users who also chat. If you already hit Codex limits, chat now competes for the same budget. A day of long agent runs can leave little for research questions in the evening.
Chat-only users on Plus. Your everyday questions may be fine in Instant mode. The change matters when you rely on advanced reasoning models for long documents or difficult analysis, because those conversations now count against the shared allowance.
Teams on Business. Check whether your workflows assume chat is separate from Codex. Budgeting by "messages" stops being meaningful once everything is counted in tokens.
Three ways to keep working when the shared pool runs low
- Move quick questions to Instant. It is outside the shared allowance, so use it for lookups, rewording and simple explanations, and keep the pool for coding and hard reasoning.
- Protect your coding budget. Start agent tasks with a narrow scope, the failing test and the exact files. A vague request burns tokens exploring the repository.
- Use a second chat service for non-coding work. Codex's value is that it runs against your repository; a browser chat cannot replace that. Long reading, drafting and comparison work, however, do not need your repository. Moving them elsewhere frees the OpenAI pool for Codex.
For the third option, AIWITH.CHAT puts GPT-6 Luna, GPT-6.1 Sol and GPT-6 Astra alongside Gemini 3.1 Pro, Claude and DeepSeek models in one account. New accounts get free trial credits without a credit card; paid plans start at $9.9/month. It has its own allowance and model-dependent costs: it is not unlimited, and it does not connect to your Codex sessions or your OpenAI account balance.
Also on December 1: Gemini 2.5 Pro and 2.5 Flash are being retired in the Gemini API
Google's Gemini API changelog (November 2, 2026) marks both models as deprecated. They can no longer be selected in Google AI Studio, and from December 1, 2026 API requests are routed automatically:
gemini-2.5-pro→gemini-3.8-flash, billed at Gemini 3.8 Flash ratesgemini-2.5-flash→gemini-3.5-flash-lite, billed at Gemini 3.5 Flash-Lite rates
Google lists breaking changes for the new models: function calling and thought_signature validation, thinking_level replacing thinking_budget, removed sampling parameters, and no prefilled model turns. If your code still names a 2.5 model, update the model string and test before December 1 rather than relying on the automatic route. Code that sets thinking_budget or prefills a model turn is the most likely to break.
If you used Gemini 2.5 Pro in AI Studio for chat, rather than through the API, you can keep using a current Gemini model in the Gemini chat on AIWITH.CHAT, which offers Gemini 3.1 Pro and Gemini 3.6 Flash.
A short December checklist
- First week of December: open ChatGPT Settings → Usage and note how much a typical day consumes.
- Route quick questions to Instant; keep advanced models and Codex for work that needs them.
- Search your code for
gemini-2.5-proandgemini-2.5-flash, switch the model string, and run your tests. - If you need a separate budget for reading, writing and model comparison, try a task on AIWITH.CHAT's GPT page with the free credits first.