codex

Codex Pricing: What the Free Tier Actually Covers

2026-08-21 · 10 min read · HabniAs of Aug 2026

Can you use Codex for free? The documentation says yes, and then declines to answer the question you actually have.

Here is the situation in one paragraph. OpenAI's pricing docs state in prose that Codex is included on Free, Go, Plus, Pro, Business, and Enterprise. But the CLI is only listed as a feature starting from the Plus card, and the usage tables and feature comparison have no columns for Free or Go at all. So the tier exists, and its boundary is undocumented.

This article maps that boundary as it actually is, separating what the docs confirm from what they leave open. Pricing has a short shelf life, so check the official pricing docs once more before you pay.

Codex: OpenAI's AI coding tool. There is a cloud version you use in a browser and Codex CLI, which you install in your own terminal. The plans cover them differently.

Codex plans at a glance: Free $0, Go $8, Plus $20, Pro from $100

As of August 2026.

Plan	Price	Is the CLI named on the official card
Free	$0	Not named
Go	$8/month	Not named
Plus	$20/month	Named
Pro	From $100/month (5x), $200 (20x)	Named
Business	$20/user/month (2+ users, billed annually)	Named
Enterprise / Edu	Contact sales	Named
API key	Pay as you go	Named

The Go plan is the row most people miss. At $8 a month it is under half of Plus. The documentation positions Go as using Codex for lighter coding work. If Plus feels steep but free is not enough, there is a middle option.

Note the last row too. It comes back later, and it is the cleanest answer in this article.

The free tier's real boundary: OpenAI documented the tier but not the CLI

This is the core of the article. Let us split what the docs say from what they do not say.

What the docs say

Codex is included on the Free plan (stated in prose)
The Free plan is for sampling Codex with quick coding tasks (plan card description)
Image generation is not available on Free (called out explicitly as an exception)

That last item matters. When a feature is unavailable on Free, the documentation says so. For the CLI, no such sentence exists.

What the docs do not say

Whether you can sign in to the CLI on a Free plan. The Plus card reads "Codex on web, CLI, IDE extension, and iOS," while the Free and Go cards carry no feature list at all
What the usage allowance is on Free or Go. The usage tables only have tabs for Plus, Pro, Business, and Enterprise
The feature comparison table also starts at Plus. There are no Free or Go columns

The honest reading is this. There is no basis to declare that Free cannot use the CLI, and no basis to promise that it can. OpenAI simply did not address that combination in its documentation.

So if you want to start free, the fastest path is to just try it. Installation is one command and it is easy to undo. Install, attempt to sign in, and judge from there. Reading more documentation will not resolve this one.

Actual usage on Plus: OpenAI publishes message counts, in ranges

One thing sets OpenAI apart from its competitors. It publishes real message counts.

Local messages available within a five hour window on Plus.

Model	Local messages per 5 hours
GPT-5.6 Sol	10 to 100
GPT-5.6 Terra	25 to 200
GPT-5.6 Luna	250 to 2,000

Pro is 5x or 20x these figures.

The ranges are wide for a reason. The documentation notes that consumption per request varies substantially with task size and complexity, model choice, context volume, and tool use. A short prompt is not automatically a cheap one. Editing a small script uses a fraction of your limit, while a long session on a large project can consume far more in a single go.

The docs also characterize the three models. Sol is for the hardest work, Terra is the daily driver, and Luna is for fast, high volume tasks.

Worth one comparison here. Anthropic does not publish concrete usage figures for Claude Code. OpenAI at least gives you ranges. Neither approach is inherently right, but if you have to budget, numbers are easier to work with than no numbers.

One more thing to know. Usage runs on a five hour rolling window, and a weekly limit may apply on top of it. Local messages and cloud chat share the same window. Heavy Codex use in the browser leaves you less in the terminal.

To see where you stand, run /status inside the CLI. It is also visible on the usage dashboard.

When you hit the limit, the docs give you four options

First, your current task finishes. The documentation states that if you reach the limit mid task, you are allowed to complete that turn. Work does not get cut off partway.

Second, drop to a smaller model. Going from Sol to Terra, or Terra to Luna, stretches the same limit further. The docs recommend this directly.

Third, buy credits. Plus and Pro can purchase additional credits and continue without upgrading the plan.

Fourth, use an API key. That gets its own section below.

Using Codex CLI without a subscription: the API key path

This is the last row from the table above, and it is the definite answer when the free tier is ambiguous.

Signing in with an API key lets you use Codex CLI. The API key card in the official pricing docs says so: you can use Codex in the CLI, SDK, and IDE extension, paying API rates for exactly the tokens Codex consumes.

The exclusions are stated too. Cloud features like GitHub code review and Slack integration are not included. Available models follow whatever that key can access.

This also mixes with other plans. The documentation notes that any user can run additional local chats with an API key, billed at API rates.

Here is the order I would use. Install for free and see whether sign-in works. If it does, use it for a few days and watch when the limit hits. If you hit it daily, move up to Go or Plus. If you only reach for it occasionally, an API key is cheaper than a subscription. Whether a fixed $20 a month or pay as you go wins is decided by how often you actually use it.

Which Codex plan to pick, by how you work

Trying it out for the first time: install on Free and see how far you get
Occasional use, a few times a month: API key, pay as you go
Light daily coding: Go at $8
Daily driver, hitting limits regularly: Plus at $20
Long sessions on large projects: Pro
A team of two or more: Business

In 30 seconds

OpenAI's docs state in prose that Codex is included on the Free plan, but the CLI is only named as a feature from the Plus card up.
No usage figures are published for Free or Go. The usage tables and feature comparison both start at Plus.
There is no documented basis either to rule out or to guarantee CLI access on Free. Installing and attempting sign-in is faster than more reading.
The Go plan at $8 a month is the middle option most people overlook.
On Plus, a five hour window allows roughly 10 to 100 messages on Sol, 25 to 200 on Terra, and 250 to 2,000 on Luna. Pro is 5x or 20x that.
Local CLI messages and browser cloud chat share the same window.
Signing in with an API key gives you CLI access without a subscription, paying only for tokens used. Cloud features like GitHub review and Slack are excluded.

Frequently asked questions

Is Codex free to use?

The pricing documentation states in prose that Codex is included on the Free plan. However, the CLI is only listed as a feature starting from the Plus card, and no usage figures are published for Free. Whether CLI sign-in works on a free account is not documented either way, so the practical answer is to install it and try signing in.

Can I use Codex CLI without a subscription?

Yes, through an API key. The official pricing docs state that with an API key you can use Codex in the CLI, SDK, and IDE extension, paying API rates for the tokens consumed. Cloud features such as GitHub code review and Slack integration are excluded, and available models follow what the key can access.

How much can I use on Codex Plus?

Within a five hour window, roughly 10 to 100 local messages on GPT-5.6 Sol, 25 to 200 on Terra, and 250 to 2,000 on Luna. The ranges are wide because consumption varies with task size, model choice, context volume, and tool use. Pro is 5x or 20x these figures.

What is the Go plan?

An $8 per month tier that OpenAI positions for lighter coding work with Codex. It sits between Free and the $20 Plus plan. Like Free, its card carries no feature list and no published usage figures, so CLI availability is undocumented there as well.

What happens when I hit my Codex limit?

The documentation lists four options. Your current turn is allowed to finish rather than being cut off. You can drop to a smaller model to stretch the same limit. Plus and Pro can buy additional credits without upgrading. Or you can sign in with an API key and pay per token.

Does browser Codex use share the same limit as the CLI?

Yes. Local messages and cloud chat draw from the same five hour window, with a weekly limit potentially applying on top. Heavy use in the browser directly reduces what remains for the terminal. Run /status in the CLI to see current standing.

Sources (5)Expand to see all sources

#Codex#Codex CLI#OpenAI#pricing#free tier