Codex “usage” is tracked in two totally different places depending on how you logged in, and that’s why your OpenAI Platform → Usage page stays at $0.
1) Two ways to use Codex (and they are metered differently)
A) Codex logged in with ChatGPT account (subscription mode)
- You sign in to Codex (web/CLI/IDE) using your ChatGPT login
- You consume message-based limits (5-hour window + weekly caps), and possibly ChatGPT credits when you exceed included limits (OpenAI Developers)
- This usage is NOT shown on platform.openai.com/usage (because it’s not API-key billing)
B) Codex using an API key (API billing mode)
- You use an API key (SDK/CLI/your app)
- You pay per token (usage-based billing) (OpenAI Developers)
- This usage IS shown on platform.openai.com → Usage (tokens/requests/spend)
That’s why your screenshot shows $0, 0 tokens, 0 requests—it only reflects API-key traffic, and you haven’t sent any. (OpenAI Developers)
2) How to track usage (exactly)
If you’re using Codex via ChatGPT login (the one giving “rate limit resets on…”)
Track it here: Codex Usage Dashboard (OpenAI Developers)
In a browser (while logged into ChatGPT), open:
https://chatgpt.com/codex/settings/usage
That page is where OpenAI says you can see your current limits. (OpenAI Developers)
While you’re inside Codex CLI, you can also see remaining limits with:
/status(OpenAI Developers)
If you’re using Codex via API key
Go to platform.openai.com → Usage (the page in your screenshot). That will show tokens/spend only after you make API requests. (OpenAI Developers)
3) Quick way to tell which mode you’re in
- If you see “You’re out of Codex messages… resets on …” → you’re hitting ChatGPT subscription limits (track in Codex usage dashboard /
/status). (OpenAI Developers) - If your Platform Usage page shows tokens/spend increasing → you’re using API key billing. (OpenAI Developers)
This is exactly the clarification I needed regarding Codex billing visibility. It’s very easy to get confused when the OpenAI platform usage stays at $0 while actively using the tool, so breaking down the difference between the ChatGPT subscription model (message-based) and the API key model (token-based) is incredibly helpful. The tip about checking the specific Codex usage dashboard versus the standard platform dashboard is a great takeaway for anyone trying to monitor their limits effectively.