Codex Usage Limit Reached: What It Means and How to Keep Working
Key takeaways
- Codex on a ChatGPT plan can have a five-hour window and a weekly window, and you need allowance left in both (as of November 2026).
- Pro plans currently have no five-hour limit.
- Settings → Usage shows which limit you reached and when it resets.
- Depending on your account, you can wait, add credits, apply a reset, or upgrade.
This was checked against OpenAI's published pages in November 2026. The rules have moved several times over the past year, so confirm them on Codex pricing and Using Codex with your ChatGPT plan.
What the message looks like
OpenAI does not publish the full text. A Codex CLI v0.166.0 user reported this on GitHub in November 2026:
You've hit your usage limit. Upgrade to Pro (https://chatgpt.com/explore/pro), visit https://chatgpt.com/settings/usage to purchase more credits or try again at Nov 4th, 2026 12:19 AM.
Older reports give a relative wait, such as try again in 4 days 2 hours 46 minutes. The wording varies by plan and version, and the time at the end is what you need.
Some GitHub reports describe this message appearing while /status still shows allowance. If that happens, compare the CLI with the Usage page in a browser, and contact OpenAI support if it continues.
How the windows work
| Item | Details |
|---|---|
| Windows | Five-hour and weekly |
| To keep working | Allowance left in both |
| What drives usage | Model, task size and complexity, local or cloud |
| Pro plans | No five-hour limit at present |
| Where to check | Settings → Usage |
OpenAI says prompt length alone is not a reliable estimate, and cloud tasks may use more than local messages. You can run out of the five-hour window with weekly allowance to spare. ChatGPT Work and Codex share usage.
Ways to keep working
OpenAI's help center says Codex can finish the current turn if you reach a limit mid-task. After that, the Usage page or the limit banner lists your options:
- Wait for the reset.
- Add credits (eligible Plus and Pro accounts).
- Apply a saved or purchased reset (depends on account and billing country).
- Upgrade your plan.
OpenAI support does not reset limits. It can investigate if usage looks miscounted or access has not returned after the stated time.
Run Codex CLI on a FastMetal key
Codex CLI lets you choose its model provider in a config file. FastMetal is an AI gateway with 95+ models behind one API key, prepaid and billed per token, with a $5 minimum top-up and no subscription.
Put the key in an environment variable:
export FASTMETAL_API_KEY="your-fastmetal-key"
Then define the provider in ~/.codex/config.toml:
model = "gpt-5.6-terra"
model_provider = "fastmetal"
[model_providers.fastmetal]
name = "FastMetal"
base_url = "https://api.fastmetal.ai/v1"
env_key = "FASTMETAL_API_KEY"
wire_api = "responses"
We ran a real Codex task to completion with Codex v0.160.1 and gpt-5.6-terra on this configuration in October 2026. env_key takes the variable's name, never the key. The provider IDs openai, ollama and lmstudio are reserved. Per-model rates are on the pricing page.
FAQ
When does the Codex limit reset?
The limit message and Settings → Usage both show the time. The five-hour and weekly windows are counted separately, so the wait depends on which one you hit.
Does a new session clear the limit?
No. OpenAI's help center says starting a new chat does not reset limits.
Why does Codex stop when I still have allowance?
Users have reported this on GitHub, and OpenAI has not published a full explanation. Restart the CLI, compare with the Usage page, and contact support if it continues.
Does a separate API key use my plan's allowance?
Requests sent to another provider such as FastMetal do not use your ChatGPT plan's allowance. That provider's pay-as-you-go billing applies.
Next step
To keep a fallback ready for work that cannot wait, create a free account and check pricing.