AI Cooldownall guides

updated September 21, 2026

Codex usage limits, explained

Codex, whether you use the CLI, the IDE extension or the cloud agent, draws on your ChatGPT plan's limits: a five-hour window and a seven-day window, each shown as a percentage with a reset time. Plus has smaller windows than Pro, and Business and Enterprise seats have their own allowances. When a window is full you can wait for its reset, switch to another account, or buy credits if your plan offers them.

The two windows

OpenAI reports both windows through the same usage endpoint the Codex CLI reads, as a percentage used plus the exact reset time. Those are the numbers AI Cooldown shows, so they match what the CLI tells you.

  • Primary, five hours. Opens with your first Codex request and closes five hours later. This is the one that stops you mid-task.
  • Secondary, seven days. A rolling weekly cap over everything Codex does on the account. Heavy daily use runs into this one.

What counts

Every Codex task counts, weighted by how much the model works. Long agent runs with many tool calls, big repositories and large diffs cost more than a quick edit. Different models cost different amounts against the same window; the heavier reasoning models burn it faster. Plain ChatGPT chat has its own message limits and is counted separately from Codex.

Model availability and credits

Alongside the windows, the usage endpoint reports whether each model is currently available to you and, when it is not, when it comes back. A model can be temporarily unavailable on your plan even while your windows have room. AI Cooldown lists that per account.

Some plans can buy credits that are used once the included windows are exhausted. If your account has a credit balance, or unlimited credits, the endpoint reports it and the dashboard shows it as a note on the account.

How to check your Codex usage

  • In the Codex CLI, the /status command shows the current windows and resets for the signed-in account.
  • In ChatGPT, the account settings show usage for that account.
  • AI Cooldown shows every window for every Codex and Claude account you connect, with countdowns, a Go or Wait verdict per provider, and which account has the most room. Sign in through the normal OpenAI login or paste the token from ~/.codex/auth.json.

see your exact numbers

Want the exact reset for your accounts?

Connect your Claude and Codex accounts to AI Cooldown and get every window with a live countdown, which one is limiting you, and which account to use next. Free, open source, and it reads the same official usage endpoints the CLIs use.

Connect an account

Questions people ask

Does Codex share limits with ChatGPT chat?
They belong to the same plan but are counted separately. Codex has its own five-hour and weekly windows; chat messages have their own caps.
Are Codex limits the same on Plus and Pro?
No. Pro windows are considerably larger than Plus windows. Business and Enterprise seats have their own allowances set by the workspace.
What happens when the five-hour window is full?
Codex stops accepting tasks on that account until the window resets, five hours after it opened. The weekly window may still have plenty left; it is the five-hour one that is blocking you.
Can I keep working on another account?
Yes. Limits are per account. AI Cooldown shows all of them side by side so you can see which one has room.

related