Navy card with a faint dot grid and sparkle marks. The label reads AI INDUSTRY. The heading reads One limit for ChatGPT + Codex, with the line From December 2026. Three chips read Plus, $20 a month; Pro, $100 to $500 a month; and Business, shared pool. Pickurai.ai sits near the bottom above a green bar.

I code in Cursor most days, from Spain, and I keep an eye on every AI plan I pay for or might pay for. When OpenAI updated its help center to say that ChatGPT and Codex will share one usage allowance from December, the question I had was simple: will I run out sooner, and what can I do about it before it happens?

Checked 2 November 2026

  • What changes. Starting in December, Plus, Pro and Business share one usage allowance across ChatGPT and Codex, measured in tokens, as the agentic features of Work move into the main ChatGPT experience.
  • What stays out. Instant chat does not draw from the allowance. Regular Chat Voice has its own limits and does not consume Codex usage.
  • What is not published. The size of the shared allowance and the exact day in December. OpenAI says it expects most users won't need to wait for a reset more often than they do now.
  • Where to watch it. Settings → Usage will show the combined remaining usage and reset times. In the Codex CLI, /status shows what is left.

What counts against the shared limit, and what does not

The usage limits help article is short. Usage will be measured in tokens, the units a model reads and writes, and how much you spend depends on the model you choose and the task. More advanced models and complex tasks use more. Instant chat does not count.

Part of this already happens. The Codex plan page says Codex, ChatGPT Work, ChatGPT for Excel and Workspace Agents already use a shared allowance and credit pool where those features are on your plan. A task you start by voice inside Work or Codex on desktop uses that same pool, while regular Chat Voice, file uploads and ChatGPT image limits are counted separately. December adds the main ChatGPT experience to the pool.

  • Inside the limit: Codex (app, CLI, IDE extension, cloud), ChatGPT Work, ChatGPT for Excel, Workspace Agents, and from December the agentic work in main ChatGPT.
  • Outside the limit: Instant chat and regular Chat Voice.

Which model gets you more tasks

OpenAI's Codex pricing page publishes estimates of local messages per five-hour window for Plus and Standard Business. They are ranges, not fixed limits, and cloud tasks use more than local messages. Pro plans currently have no five-hour limit, though weekly limits may apply.

Model Local messages per 5 hours (Plus and Standard Business)
GPT-6 Astra5–45
GPT-6.1 Sol15–160
GPT-6 Sol15–150
GPT-6.1 Terra35–285
GPT-6 Luna350–3,000

Estimates from the ChatGPT Learn pricing page, checked 2 November 2026. Fast mode uses included usage at 2.5× the Standard rate, and Ultrafast at 8×. Image generation uses limits about 3–5× faster.

The spread is the point. On the same Plus seat, the top of the Terra range is more than six times the top of the Astra range, and Luna's is far higher again. If December means your agent work and your ChatGPT work drain the same bucket, the model picker is the first lever you have. I covered when Terra is good enough in GPT-6.1 Terra vs Sol vs Astra.

What to do, by profile

You mostly chat. Instant chat does not draw from the allowance and regular voice has its own limits, so on what is published you should notice little. OpenAI says it expects most users won't hit resets more often than now.

You have Plus and code in Codex every day. This is the profile that should plan ahead, because from December your ChatGPT agent work spends the same pool. Default to Terra or Luna for routine work, keep Astra for the hardest tasks, avoid Fast mode when you are not in a hurry, and keep prompts, files, AGENTS.md and the number of MCP servers lean, which are the tips OpenAI itself gives. Run /status before a long session.

You are on Business. Standard Business has the same five-hour estimates as Plus. Workspaces with flexible pricing can buy extra workspace credits when the pool runs out, so it is worth asking your admin which billing model you are on before December.

When you hit the limit. If it happens mid-turn, the agent can finish that turn, subject to fair use. After that you can wait for the reset, buy credits (Plus and Pro can do this without changing plan), upgrade, or run extra local chats with an API key at API prices. Plus is $20 a month, and Pro comes at $100, $200 or $500. Whether credits, an API key or Pro is cheaper depends on how often you run out; API vs subscription and the token calculator help you put numbers on your own usage.

How to watch your usage

Settings → Usage will show combined remaining usage and reset times. In an active Codex CLI session, /status shows your remaining limits. OpenAI suggests checking the dashboard every week or two to see your pace. Support does not reset usage limits, so if something looks wrong, OpenAI asks you to contact them with the balance, reset time, client and model.

I will update this page when OpenAI publishes the date or the size of the shared allowance. If you want the broader picture of trimming AI plans, how I manage my AI plans and plans that let you choose the model are the longer reads.

Sources: