Codex usage limits, explained
Codex comes with a ChatGPT plan, and its usage is metered in a five-hour window with a weekly limit behind it. How much you get depends on the plan, the model and where the task runs.
Short answer. ChatGPT Plus and Standard Business have a five-hour Codex limit; Pro plans currently don’t. Weekly limits may apply to all of them. On Plus, OpenAI estimates 15–150 local messages per five hours on GPT-6 Sol and 350–3,000 on GPT-6 Luna. Check what’s left with /status in the CLI.
Limits by plan
| Plan | Price | Five-hour limit | Weekly |
|---|---|---|---|
| Free, Go | $0, $8/mo | Not published; GPT-6 Luna only | Not published |
| Plus | $20/mo | Yes | May apply |
| Pro | $100, $200 or $500/mo | None at the moment | May apply |
| Business | $20/user/mo yearly, $25 monthly | Yes on Standard, same as Plus | May apply |
| Enterprise, Edu | Contact sales | Flexible pricing: no fixed limits, usage scales with credits | — |
How many messages in five hours
OpenAI publishes estimates of local messages per five-hour window for Plus and Standard Business. They are ranges because a message’s cost depends on context, reasoning, tool use and caching.
| Model | Local messages per 5 hours |
|---|---|
| GPT-6 Astra | 5–45 |
| GPT-6 Sol | 15–150 |
| GPT-6.1 Sol | 15–160 |
| GPT-6 Luna | 350–3,000 |
What uses more
- Cloud tasks may use more of your allowance than local messages; both share it.
- Fast mode uses included usage 2.5 times faster; GPT-6 Astra Ultrafast 8 times.
- Higher reasoning effort uses more tokens and takes longer.
- Code review through GitHub has its own allowance; reviews run locally count toward your general limit.
How to check your Codex usage
/statusin the Codex CLI shows the chat ID, context usage and your rate limits./usageshows token usage (daily,weeklyorcumulative) and lets you redeem an earned rate-limit reset, if you have one./statuslinecan keep the limits in the CLI footer.- On the web:
chatgpt.com/codex/settings/usage.
When you hit the limit
The CLI says You’ve hit your usage limit, with an offer to upgrade or buy credits, “or try again at” a time. If the limit applies to one model, the message says so and suggests switching. The turn already running can finish.
- Wait for the time in the message.
- Buy credits on Plus or Pro; OpenAI lists credits per million tokens for each model.
- Redeem a reset with
/usage, if you have earned one. - Use an API key for extra local chats at API rates.
- Switch agents. If you also have Claude Code or OpenCode Go, here is how to hand the task over.
Make it last longer
From OpenAI’s own advice: keep prompts tight and give only the source material the task needs, ask for output no longer than you need, keep AGENTS.md small by nesting it in subfolders, use fewer MCP servers, and pick a smaller model for routine work.
Codex limits in your MacBook notch
CapaTheNotch shows what’s left in the Codex five-hour and weekly windows and when they reset, next to Claude Code or OpenCode Go. It asks Codex through its own App Server, so your credentials never leave Codex. Free and open source.
Download for macOSSources
- ChatGPT and Codex pricing (OpenAI)
- Codex manual (OpenAI)
- openai/codex, for the CLI’s limit messages
FAQ
How many Codex messages do I get on ChatGPT Plus?
OpenAI estimates local messages per five hours by model: 5–45 on GPT-6 Astra, 15–150 on GPT-6 Sol, 15–160 on GPT-6.1 Sol and 350–3,000 on GPT-6 Luna. These are estimates, not fixed limits.
Does ChatGPT Pro have a Codex limit?
Pro plans currently have no five-hour limit. Weekly limits may still apply.
How do I check my Codex usage?
Type /status in the Codex CLI to see your rate limits, use /usage for token usage, or open chatgpt.com/codex/settings/usage.
What happens when I hit the Codex limit?
The current turn can finish. After that you can wait for the reset time in the message, buy credits on Plus or Pro, switch to a cheaper model if the limit is model-specific, or use an API key at API rates.
Do cloud tasks use more Codex usage than local ones?
Yes. OpenAI says cloud tasks may use more of your allowance than local messages; both draw on the same allowance.