ChatGPT Codex Shared Usage Limit: What Changes in December

The ChatGPT Codex shared usage limit starts in December 2026: Plus, Pro, and Business plans will draw ChatGPT and Codex usage from one shared allowance measured in tokens. Instant chat will not count against it. OpenAI has not yet published the exact start date or how many tokens each plan gets, so the safest move now is to learn how your current usage burns and where to watch the new combined meter.

Last verified: November 2, 2026. Based on OpenAI Help Center and ChatGPT Learn pages as they read on that date. This is a pre-launch article: the change takes effect in December 2026, and details may change before or after rollout.

What is the ChatGPT Codex shared usage limit starting in December?

From December 2026, Plus, Pro, and Business subscribers get one usage allowance that covers both ChatGPT and Codex. Usage is counted in tokens, Instant chat is excluded, and the combined balance and reset times will appear under Settings > Usage.

OpenAI announced the change in its Help Center article “Understanding ChatGPT and Codex usage limits,” which several outlets reported was updated on October 31, 2026. The company ties the change to bringing the agentic features of ChatGPT Work into the main ChatGPT experience. In plain terms, the line between “chatting” and “handing off a task to an agent” is getting thinner, so OpenAI is moving both onto one meter.

The first public hint came a day earlier. On October 30, product tracker Tibor Blaho spotted a line on the ChatGPT checkout page saying that, starting in December, plan usage would share one limit across ChatGPT and Codex. The Help Center update then confirmed the plan for three tiers.

This is not a brand-new idea inside OpenAI’s products. Codex, ChatGPT Work, ChatGPT for Excel, and Workspace Agents already draw from one shared agentic pool on plans where those features are available. The December change mainly pulls everyday ChatGPT use into that same accounting.

What counts against the ChatGPT Codex shared usage limit?

Under the new model, ChatGPT and Codex usage on Plus, Pro, and Business count against one token allowance, with Instant chat as the only exclusion OpenAI has named. Codex use from the web, the CLI, and the IDE extension already shares one pool today.

Here is how the pieces fit, based on what OpenAI has published so far:

• Codex (all surfaces): OpenAI’s pricing page says local messages and cloud chats share your plan’s usage allowance. That covers the Codex CLI, the IDE extension, the desktop app, and Codex Cloud when you sign in with ChatGPT.
• ChatGPT Work, ChatGPT for Excel, and Workspace Agents: These already use the same agentic pool as Codex on plans where they are available.
• Everyday ChatGPT (non-Instant): This is the new part. From December, it draws from the same allowance.
• Instant chat: OpenAI says it will not draw from your usage. OpenAI has not said Instant becomes unlimited, only that it sits outside this allowance.
• API-key usage: Codex sessions authenticated with an API key are billed at standard API rates, not from your ChatGPT plan.

Some ChatGPT features have historically run on their own caps. OpenAI’s Codex help page says ChatGPT file uploads, image generation, and voice have separate usage limits, reset periods, and banners, and that ordinary Chat Voice does not consume Codex usage. Whether all of these stay separate after December is not spelled out in the new article, so treat that as an open question until OpenAI clarifies.

What changes for ChatGPT Plus, Pro, and Business subscribers?

All three plans move to one ChatGPT and Codex allowance. Plus and Standard Business currently have a five-hour window plus possible weekly limits, while Pro plans currently have no five-hour limit. OpenAI has not said how those windows change under the shared model.

The table below uses prices and limit rules from OpenAI’s ChatGPT Learn pricing page, checked on November 2, 2026. Prices are in USD per month.

How to read this table: the “Shared allowance” column reflects only the three plans OpenAI names. For Free, Go, Enterprise, and Edu, the article does not say, so check your own account rather than assuming either way. The time-window column describes the current system, not the December terms.

Plus: the plan most likely to feel the squeeze

Plus users have the smallest allowance among paid personal plans and a five-hour window that OpenAI restored for Work and Codex in late August 2026. Under one pool, a long Codex session in the afternoon can reduce what is left for ChatGPT that evening. The reverse also holds: allowance you did not use in ChatGPT can now cover extra Codex work.

Pro: more headroom, same accounting

Pro users still get one combined balance, but their current Work and Codex usage has no five-hour cap. OpenAI’s Pro help page notes that the $100 tier includes lower allowances than the $200 tier. In late October, reports said OpenAI doubled allowances for several Pro tiers and Business Premium; Plus was not part of that increase.

Business: one number to forecast

For teams, the shared pool replaces two balances that could run dry at different times. That makes per-seat budgeting simpler, but heavy Codex users on a team may notice their chat headroom shrink after a big coding sprint.

How does token-based usage work for ChatGPT and Codex?

Tokens are small units of text the model reads and writes. Your prompt, attached files, chat history, tool results, and the model’s reply all use tokens, so a long agentic task can use far more of your allowance than a short question, even if both are “one message.”

OpenAI’s pricing page lists what drives usage: the model you pick, the size and complexity of the task, how much context the agent holds, reasoning, tool calls, retrieval, caching, and whether the work runs locally or in the cloud. Cloud tasks may use more of your allowance than local messages.

Settings that change how fast the pool drains

These multipliers come from OpenAI’s pricing page as of November 2, 2026. They describe how fast usage is consumed relative to Standard speed for the same model. They do not tell you the size of any plan’s allowance.

Model choice matters more than prompt length

OpenAI publishes rough local-message estimates per five-hour window for Plus. For example, GPT-6.1 Sol is listed at 15–160 local messages, while the lighter GPT-6 Luna is listed at 350–3,000. The wide ranges show why message counts are a poor budget tool: the same model can cost very different amounts depending on the task. OpenAI itself says prompt length alone is not a reliable estimate.

What happens when you hit the shared ChatGPT and Codex limit?

You have three options: wait for the reset, upgrade to a higher plan, or buy credits for eligible features. OpenAI says it expects most users will not have to wait for resets more often than they do now, but that is a forecast, not a guarantee.

1. Wait for the reset. Settings > Usage will show remaining usage and reset times for the combined allowance.
2. Upgrade your plan. Moving from Plus to a Pro tier raises the allowance. OpenAI says new Pro limits apply as soon as you switch.
3. Buy credits. On Plus and Pro, credits let you keep using supported features after included usage runs out.

How credits work today, according to OpenAI’s personal-plan credits article:

• Your plan’s included usage is used first. Credits are drawn only after you hit plan limits.
• You can buy credits in Settings > Usage in ChatGPT or Usage & Billing in the Codex app. Automatic reload is available on some accounts.
• Credits are valid for 12 months from purchase and do not roll over after they expire.
• Buying credits does not raise your included plan limit. It adds a separate pay-as-you-go balance.

OpenAI also notes that if you hit a limit during an active turn, the agent can usually finish that turn, subject to fair-use limits. And OpenAI support does not manually reset ChatGPT or Codex limits, per its Pro help page.

How to prepare your ChatGPT and Codex workflow before December

Measure your current usage now, cut avoidable context, and decide which tasks need a heavy model. A few weeks of data from the usage dashboard will tell you far more than any guess about December allowances.

Common mistakes to avoid

• Assuming message counts still matter. The new meter is token-based, so one long agent run can outweigh dozens of short chats.
• Buying credits before you know your pattern. Credits expire after 12 months. Measure first, then buy.
• Trusting third-party “token amount” claims. As of November 2, 2026, OpenAI has published no per-plan token figures.
• Forgetting that cloud tasks cost more. OpenAI says Codex Cloud tasks may use more allowance than local messages.

What has OpenAI not announced yet about the shared limit?

OpenAI has not published the exact December start date, the token allowance for each plan, or how the five-hour and weekly windows will work under the shared model. Until it does, any specific number you see elsewhere is a guess.

OpenAI’s own pages also do not fully line up yet. The Codex plan help page and the Learn pricing page still describe the current system, including Pro’s lack of a five-hour limit and Plus local-message ranges. Expect those pages to be revised as the December rollout gets closer.