Which plans include Codex
Codex is part of ChatGPT Plus ($20/month), the three Pro plans ($100, $200 and $500/month) and Business, on the web, in the CLI, in the IDE extension and in the apps. Free and Go get a lighter version in the desktop app as it rolls out, without cloud tasks (Codex pricing, Using Codex with your ChatGPT plan).
Two windows: 5 hours and a week
Like Claude, Codex usage runs on two windows at once: a 5-hour window and a weekly window. The CLI shows both as "5h limit" and "Weekly limit", each with its reset time. Two details matter:
- Pro plans currently have no 5-hour limit, only the weekly one (OpenAI).
- Everything shares one allowance. Local work in the CLI or IDE and cloud tasks draw from the same plan allowance, and cloud tasks can use more of it.
OpenAI publishes estimated message ranges per 5 hours for Plus and Business, and says they are "not fixed message limits". How far your allowance goes depends on the model and the task.
How to check your Codex usage
- On the web: chatgpt.com/codex/settings/usage shows your usage and reset times.
- In the Codex CLI: run
/status./usage daily,/usage weeklyand/usage cumulativebreak it down (CLI commands). The CLI also warns you when you're down to 25% and 5%. - Always in view: Usage for Claude + ChatGPT shows your Codex 5-hour and weekly limits in the Mac menu bar, on your iPhone and in widgets, next to your Claude accounts, with alerts before you run out.
When you hit the limit
If you run out in the middle of a turn, Codex can usually finish that turn. After that, your options are (OpenAI, credits):
- Wait for the window to reset.
- Use credits. Plus and Pro can buy credits under Settings › Usage in ChatGPT. Your included usage is spent first, then credits. They last 12 months and aren't refundable.
- Use a banked reset, if a promotion gave you one. It refreshes both windows.
- Upgrade your plan.
- Sign in with an API key for local work, billed at API rates.
OpenAI support can't reset your limits for you.
What uses it up faster
OpenAI lists model choice, context, reasoning effort, tool use, retrieval and caching as the main factors (Codex pricing). In practice:
- The model. A small model lasts many times longer than the largest one.
- Reasoning effort. Higher effort uses more tokens.
- Speed modes. Fast mode uses your allowance at 2.5x, and Ultrafast faster still (Pro tiers).
- Cloud tasks can cost more than the same work run locally.
- Context you add to every turn: each MCP server, a large AGENTS.md and long sessions all add up.