OpenAI Codex does not have one single monthly price. Instead, it comes bundled inside ChatGPT plans, including Free, Go, Plus, Pro, Business, Edu, and Enterprise. The most popular options are Plus at $20 per month, Pro 5x at $100 per month, and Pro 20x at $200 per month. You can also pay through an API key, where you only pay for the tokens you use.
That is the short answer. The full picture needs a little more detail, because OpenAI changed how Codex billing works in April 2026. This guide explains every plan, how credits work, and which option fits your budget.
What Is OpenAI Codex?
Before we talk money, let us clear up one common mix up. There are two products called OpenAI Codex. The original Codex from 2021 was a code model that powered early GitHub Copilot, and OpenAI retired it in 2023. The current Codex, launched in May 2025, is a cloud based coding agent. You give it a task, it works on its own in a safe sandbox, and it comes back with finished code, logs, and test results.
Today Codex runs on the web, in a VS Code extension, in the command line, on iOS, and even on Amazon Bedrock. It is one of the most popular AI coding agents in the world, alongside tools like Claude Code and Cursor. If you are curious how agents like this work behind the scenes, our team explains it in detail on our AI agent development page.
How Much Does OpenAI Codex Cost?
Here is a simple table of every Codex plan in 2026.
| Plan | Price | Best For |
|---|---|---|
| Free | $0 | Trying Codex on small tasks |
| Go | Low cost entry plan | Light coding help |
| Plus | $20 per month | Hobby coders and part time developers |
| Pro 5x | $100 per month | Full time developers |
| Pro 20x | $200 per month | Heavy users and engineering leads |
| Business | $25 per user per month billed monthly, lower when billed annually | Startups and teams |
| Enterprise | Custom pricing | Large companies |
| API Key | Pay per token | Automation and CI/CD pipelines |
A few important notes on these plans.
Plus at $20 per month is the practical starting point for working developers. It includes cloud task delegation, GitHub code review, and Slack integration.
Pro comes in two levels. The $100 tier gives you 5 times more Codex usage than Plus. The $200 tier gives you 20 times more. OpenAI added the Pro 5x tier in April 2026, partly to match Claude Code at the same price point.
Business plans offer two seat types. Standard seats include full ChatGPT plus Codex at a fixed price per user. Codex only seats give access to Codex alone and are billed by usage. This is handy because a project manager and a build server do not need the same seat.
How Does the New Credit System Work?
This is the part most older articles get wrong. On April 2, 2026, OpenAI switched Codex from per message billing to token based credit billing. The change reached all Enterprise plans on April 23, 2026.
Here is the plain English version. Every task you give Codex uses tokens. Tokens are small chunks of text that the AI reads and writes. Codex converts your token usage into credits based on three things: input tokens, cached input tokens, and output tokens.
A typical Codex task using GPT-5.5 may use between 5 and 45 credits. That is a wide range, and it explains why two developers on the same plan can have very different bills. A complex task that touches many files can burn around 9 times more credits than a simple script edit.
Your usage also refills on a rolling 5 hour window, not a monthly bucket. Heavier models and longer agent runs consume more of that window.
Three things push your credit usage up the most:
- Model choice. GPT-5.3-Codex has a higher credit rate per token, while GPT-5.4-Mini is cheaper and made for lighter tasks.
- Fast mode. Fast mode burns credits at a higher rate on supported models.
- Task size. Big refactors across a large codebase cost far more than a quick bug fix.
You can watch your spending inside the app. Open Codex settings and check the Usage panel to see your limits, remaining credit, and options to buy more.
What Do Developers Actually Pay Per Month?
Sticker prices and real costs are different things. OpenAI’s own rate card estimates that active users spend around $100 to $200 per developer per month, with big swings based on the model used, the number of agents running at once, automations, and fast mode.
The same subscription tier can cost one developer $40 and another $400, depending on how they work. So here is a simple way to pick:
- You code as a hobby or part time: Plus at $20 is enough. Most hobby developers never outgrow it.
- You code full time and hit Plus limits often: Pro 5x at $100 is the sweet spot.
- You run parallel agents all day or automate code review at scale: Pro 20x at $200.
- You have a team: Business, with a mix of standard and Codex only seats.
Is the API Cheaper Than a Subscription?
For most people, no. Running about 200 cloud tasks a month through the API would cost roughly $50 to $90, which is 2.5 to 4.5 times the price of a Plus subscription. The flat $20 plan quietly rewards moderate users.
The API wins in different situations. It is the best choice for CI/CD jobs, programmatic workflows, SDK usage, and automation in shared engineering environments. For current per token rates, always check the official OpenAI Codex pricing page, since rates change often.
One warning: API access skips some cloud features like GitHub code review and Slack integration, and new models often reach subscribers before API users.
If your company wants to build custom tools on top of the OpenAI API and needs help planning the costs, that is exactly the kind of work we handle in our AI consulting and strategy service.
How Does Codex Pricing Compare to Cursor and Claude Code?
At $20 per month, Codex Plus and Cursor Pro cost the same but work very differently. Codex runs tasks in the background while you do other things. Cursor works with you in real time inside your editor. Many developers who use both say Codex is stronger for complex backend tasks, while Cursor feels better for interactive frontend work.
At the $100 level, Codex Pro 5x competes directly with Claude Code. Plenty of developers actually run both, sending hard backend problems to Codex and iterative work to Claude Code.
We have covered these tools before on our blog. If you are choosing between editors, read our comparison of Augment Code vs Cursor and our honest guide on Cursor AI for model development.
Frequently Asked Questions
Is OpenAI Codex free to use?
Yes. The Free plan offers limited trial access at $0 with no credit card required. It works for small local tasks, but you will hit the wall quickly on real projects.
Why did my Codex bill change in 2026?
OpenAI moved from per message billing to token based credit billing in April 2026. Your cost now depends on how many tokens each task consumes, not how many messages you send.
How many tasks can I run on the Plus plan?
There is no fixed number. Your allowance depends on the model, the size of your coding tasks, and whether work runs locally or in the cloud. Small scripts use only a small slice of your limit, while large codebases use much more.
What is the difference between Pro 5x and Pro 20x?
Both are Pro plans. The $100 tier gives 5 times the usage of Plus, and the $200 tier gives 20 times the usage of Plus. Pick 5x if Codex is part of your daily work. Pick 20x if it runs almost nonstop.
Should my business use Codex or build a custom AI coding workflow?
Codex is great for general coding help. But if your team has repeat tasks that follow the same pattern, a custom workflow built around your own tools and data often costs less over time and fits your process better. You can explore what that looks like on our custom AI development and AI automation and workflow integration pages.
