How to Check Codex Usage in ChatGPT (Plus, Pro and Credits)

Summary: ChatGPT’s Codex has its own usage analytics page at https://chatgpt.com/codex/cloud/settings/analytics, separate from the main ChatGPT settings. You need to be logged into your ChatGPT account first. The page splits your Codex usage into three meters: a 5-hour rolling limit, a weekly limit, and pay-as-you-go credits. Here is an example screenshot:

Where EXACTLY Do I Find My Codex Usage on ChatGPT?

You have to be logged into your ChatGPT account first. Then paste this URL into your browser:

https://chatgpt.com/codex/cloud/settings/analytics

Two things worth knowing before you go looking for it:

1. This page is not under the main ChatGPT Settings menu. It is on the Codex side of ChatGPT, which is a separate area of the product.
2. The fastest way to land here when you are locked out is to click the red “Usage limit reached” pill in the Codex header. ChatGPT routes you straight to the analytics page.

If you work mostly in the terminal, the Codex CLI can pin these meters to a status line so you never have to open the analytics page. Here is my Codex CLI status line walkthrough.

What Are the Three Meters?

Three independent budgets are tracked on this page:

5-hour usage limit. This is the one most people hit first. Rolling window. Local Codex messages and cloud tasks both draw from it. The page shows the reset timing for your current window.

Weekly usage limit. Separate cap on top of the 5-hour. You can hit your 5-hour wall and still have 80%+ of your weekly intact, which is exactly what my screenshot shows.

Credits. Top-up that kicks in after you hit included Codex usage limits. Codex credit pricing has been token-based since April 2026. Plus and Pro users can buy them and the balance feeds back into the same Codex pool. Since June the Codex CLI /usage command also surfaces any earned limit-reset credits, and since July 9 it shows each credit’s type and expiration and lets you choose which one to redeem. Banked resets lapse 30 days after they are credited.

The card right on the analytics page reads: “Codex usage draws from your shared agentic usage limit.” OpenAI’s pricing docs now put it as “ChatGPT Work and Codex share usage.” Heavy Codex use eats into your ChatGPT Work headroom too.

How Many Codex Messages Does ChatGPT Plus Actually Give Me?

Per OpenAI’s official Codex pricing page, here is the September 2026 rate card for the $20/month Plus tier, per 5-hour window. These are local messages; OpenAI notes cloud tasks can draw more from your allowance than local ones:

Model	Local messages per 5-hour window
GPT-6 Astra	5 to 45
GPT-5.6 Sol	10 to 100
GPT-5.6 Terra	25 to 200
GPT-5.6 Luna	250 to 2,000

The range is wide because consumption varies by message size. A 50-line diff review costs more than a one-sentence question. Long prompts, attached files, and large code reviews burn faster than chat-style turns. OpenAI stresses the point right on the pricing page: “These estimates are not fixed message limits; check your usage dashboard for current limits and reset times.”

Is Upgrading to Pro Worth It?

For Plus users hitting the wall regularly, here is the math, again from OpenAI’s pricing page:

Tier	Price	Astra	Sol	Terra	Luna
Plus	$20/mo	5 to 45	10 to 100	25 to 200	250 to 2,000
Pro 5x	$100/mo	25 to 225	50 to 500	125 to 1,000	1,250 to 10,000
Pro 20x	$200/mo	100 to 900	200 to 2,000	500 to 4,000	5,000 to 40,000

Earlier versions of this table had “not published” holes in it because OpenAI was not saying. As of September 2026 the full grid is published for every current model, so those holes are gone.

If you read an earlier version of this post, the temporary Pro promos are gone. The Pro 2x promo expired May 31, so the $100 tier is back to the standard 5x and the $200 tier is back to 20x on the 5-hour window. A separate June referral promo ended June 24, and as of July 2026 nothing has replaced them. GPT-5.6 (Sol, Terra, and Luna) landed July 9, which is when the rate card went per-model. Then on September 3 OpenAI shipped GPT-6 Astra into Codex, filled in the Pro cells it used to leave blank, and revised most of the 5.6 ranges (Luna’s Plus ceiling jumped from 280 to 2,000). Sol also has promotional token pricing running through at least November 21, 2026.

If you blow through Plus more than once a week, Pro 5x is the obvious step. If you hit the wall once a month and can wait 5 hours, stay on Plus and treat the lockouts as forced breaks.

Reader: Did One Astra Prompt Eat My Whole 5-Hour Limit?

Two days after GPT-6 Astra landed in Codex, reader Kristian Feichtinger left a comment on this post: one Astra prompt on the Plus plan and his entire 5-hour meter was gone. I cannot verify what his prompt was doing, but I checked OpenAI’s numbers and the report is entirely plausible.

Astra sits at the top of OpenAI’s rate card and drains the shared pool far faster than the GPT-5.6 family. Plus gets roughly 5 to 45 Astra messages per window, and OpenAI flat-out says those are estimates, not fixed counts. At the low end, one heavy message can eat about a fifth of the window, so if one prompt wiped the whole meter, there was probably more going on: a window already partly spent, an unusually large job, or one prompt kicking off an agent run that sent several messages behind the scenes. Put the ranges side by side and the shape is plain: on Plus, Astra’s allowance is about half of Sol’s and roughly a fiftieth of Luna’s, whose range on the same plan is 250 to 2,000.

Speed settings can burn through it even faster. OpenAI says speed configurations “use included limits faster,” and Fast mode specifically runs a 2.5x multiplier on Astra’s standard rate. The pool is shared, so once the meter reads 0%, switching to a cheaper model will not get you around the lockout. You wait for the window to roll forward, or you spend credits.

Suggestions:

1. Check the meter before an Astra job. The analytics page above, or the CLI status line, and note your reset time.
2. Keep a 5.6 model as the daily driver. That is how I run it: Sol for routine passes, Astra saved for the hard stuff that deserves it. In the CLI you have to ask for Astra explicitly anyway (codex -m gpt-6-astra); it does not run by default.
3. Credits keep you moving while you are locked out, but Astra spends them at the top rate on the card (250 credits per million input tokens and 1,250 per million output, versus Sol’s 100 and 500). Finishing the job with Sol or Luna is cheaper. (as a last ditch effort on a time sensitive task)
4. Watch Fast mode. 2.5x on a model this heavy adds up quickly.

I ran the experiment myself while shipping this update: one mid-size Astra prompt (this article’s final review pass, roughly a 15 KB job) did not visibly move my 5-hour meter at all, 3% used before and 3% after, though the weekly meter ticked from 0% to 1%. I read the meter with two featherweight Luna pings, so those are baked into the readings too. The wide ranges cut both ways: my modest prompt barely registered, and a monster agent run is a different animal. Hat tip to Kristian for flagging this the week it changed. If Astra has eaten your window too (or surprised you by not doing that), tell me in the comments what the prompt was doing. The more data points, the better this page gets.

What Do I Do When I Hit 0% on my 5 hour block?

Three options when the 5-hour allotment is used:

1. ⏰ Wait it out. The 5-hour window rolls forward whether you are paying attention or not. For occasional Codex users this is fine.
2. 💰 Buy credits. Pay-as-you-go top-up at token-based rates. Useful for “I just need to finish this one task” moments. Earned reset credits can also be redeemed straight from the Codex CLI with /usage.
3. 💰💰 Upgrade to Pro. 5x quota for $100/mo

If you are locked out and the task can wait, option one is the cheapest. Maybe just take a nice break and have a phone alarm for your reset time, come back when the meter ticks over.

The End

I hope this helps! I try to keep this info updated but OpenAI frequently changes things, so if something looks amiss just let me know. Also, if you have questions, tips or tricks to share, feel free to comment. Thanks for visiting! 👍👍