Blog / Service updates

Codex Usage Limits Explained: Plus, Pro and Business 5-Hour Caps

2026-10-17cdnping editors4 min read

What do you do when you hit your Codex usage limit? First work out which limit you hit. Plus and standard Business plans are metered by estimated messages per 5-hour window. Pro currently has no 5-hour limit, and a weekly limit may also apply. Once you run out, you can switch to a lower-cost speed mode, buy ChatGPT credits, or switch to an API key with pay-as-you-go billing. This article is based on OpenAI's official pricing page and changelog (pages checked October 17, 2026; the October 8, 2026 changelog entry added details on Ultrafast mode and the $500 Pro tier). Prices and numbers are subject to change, so treat the official pages as authoritative.

#What are the Codex plans, prices and tiers?

According to OpenAI's pricing page, as of October 2026:

Enterprise and Edu plans require contacting sales.

#How many messages is the 5-hour limit?

OpenAI publishes a range, and it varies by model. The estimated number of local messages per 5 hours runs from single digits to thousands depending on the model. The figures on the page are estimates, not fixed message caps. Don't treat them as guarantees; your actual remaining allowance is whatever the usage dashboard shows.

Two more points are easy to miss:

• Local messages and cloud conversations draw from the same allowance.
• Cloud tasks may consume more of your allowance, and the numbers on the official page are only estimates.

#Is there a number for the weekly limit?

No. OpenAI only says that "weekly limits may also apply." The usage dashboard shows your current limits and when they reset. OpenAI hasn't published a specific figure, so any third-party claim of "N messages per week" is not an official number. For a history of reset times, see Codex usage limit reset timeline.

#How much extra allowance do Fast and Ultrafast modes use?

Relative to Standard mode, OpenAI gives these multipliers:

Ultrafast is available only on the $500 Pro tier and on eligible Enterprise and Edu plans. The changelog records the feature as launching on October 8, 2026. If your allowance is tight, the most direct fix is to go back to Standard.

#What to do when you run out

Work through this list in order:

1. Open the usage dashboard, confirm whether you hit the 5-hour limit or the weekly limit, and note the reset time.
2. Switch Fast or Ultrafast back to Standard, and run fewer cloud tasks.
3. Plus and Pro users can buy extra ChatGPT credits instead of upgrading right away. Workspaces on flexible billing, such as Business, can buy workspace credits.
4. In the CLI, SDK or IDE extension, switch to an API key and pay at API prices. For endpoint setup, see Custom API endpoints for Claude Code and Codex.

#Out of allowance, or a service outage?

Both look like "it won't work," but the fix is completely different. A usage problem usually comes with an explicit limit message, and the dashboard shows the allowance is exhausted. An outage looks like many users hitting errors or timeouts at the same time.

• First check cdnping's Codex status page and OpenAI API status to see whether there is an incident.
• To learn how to read these pages, see How to read the OpenAI and Claude status pages.
• If both look normal, check your own network. See Is ChatGPT down, or is it my network? A three-step check.

#Which tier should you pick?

There's no universal answer; it depends on how many tasks you run each day. One approach you can follow: use Plus for a week or two and record how often you hit the 5-hour limit each day. If the limit interrupts you every week, consider the $100 Pro tier. Only look at higher tiers if $100 still isn't enough. Ultrafast exists only on the $500 tier, so don't upgrade for it unless you really need it. For teams, compare Business per-seat pricing combined with credits.

#Sources

• OpenAI: Pricing (Codex usage limits), checked October 17, 2026
• OpenAI: Changelog, checked October 17, 2026