OpenAI Codex Pricing 2026: Plans, Limits and Credits

Last Updated: November 9, 2026

Summary: OpenAI Codex has no separate price tag. It comes with every ChatGPT plan, from Free and Go up to Plus ($20 a month), Pro ($100, $200 or $500 a month) and Business ($20 to $125 per seat), and what you really pay for is how much Codex you can use before you hit the five-hour limit. Once you hit it, you either buy ChatGPT credits or switch to an API key billed at standard API rates.

That makes "how much does Codex cost?" a harder question than it looks. The sticker price is your ChatGPT plan, but the real cost depends on which model you run, whether you use Fast mode and how often you go past your included usage.

This guide pulls every current number from OpenAI's own pricing page, rate card and help center into one place. It also adds something the official pages leave out: a plain conversion between Codex credits and API dollars, so you can tell when credits are the better deal and when an API key wins.

OpenAI Codex costs nothing extra on top of a ChatGPT subscription. OpenAI's help center says Codex "is included across ChatGPT plans, including Free and Go," while Codex Cloud (tasks that run in OpenAI's cloud) is limited to eligible Plus, Pro, Business, Enterprise, Healthcare and Education accounts.

The plan you choose decides which Codex surfaces and models you get and how much usage comes included. Here is the full lineup as of November 9, 2026.

Sources: Plus price from TechCrunch's April 9, 2026 report; Pro tiers from OpenAI's "What is ChatGPT Pro" page; Business seats from OpenAI's ChatGPT Business page; plan features and the API key option from OpenAI's Codex pricing page. OpenAI notes that annual billing isn't offered for Go, Plus or Pro.

OpenAI added the $100 Pro tier in April 2026 specifically for heavier Codex users, and TechCrunch quoted the company saying the tier "is designed to give developers more practical coding capacity for the money." The $500 tier followed on September 29, 2026 and now sits at the top of the range. OpenAI has adjusted Pro allowances since then and does not publish exact usage multiples for each tier on its current Pro help page. For the full ChatGPT plan picture beyond coding, see our ChatGPT pricing guide.

Codex limits are counted in a rolling five-hour window, and the number of messages you get depends heavily on the model. OpenAI's Codex pricing page gives these estimated ranges for local messages per five hours on Plus and on a Standard Business seat:

OpenAI calls these estimates rather than fixed caps. Cloud tasks can use more of your allowance than local messages, and weekly limits may also apply, though the page doesn't publish weekly figures.

The spread inside each range is wide for a reason. OpenAI's help center says usage depends on the model, where the task runs, task complexity, context, reasoning, speed and tools. A quick one-file fix and a long refactor that reads half your repository count very differently.

Two smaller details are worth knowing. Auto-review safety checks don't count toward your limits, and OpenAI Support will not reset a limit for you. Some Plus and Pro referral promotions add a one-time banked reset that refreshes both the five-hour and weekly windows, but OpenAI says these are occasional and not guaranteed.

If you also use Anthropic's tools, our breakdown of Claude usage limits shows how the two companies meter coding work differently.

Codex credits are a prepaid balance that kicks in after your plan's included usage runs out. Per OpenAI's credits help page, plan-included usage is always used first, and Plus and Pro users can buy credits without upgrading. Credits are valid for 12 months from purchase and don't roll over after that.

Credits are charged per million tokens, at a different rate for each model. These are the Standard-speed rates from OpenAI's Codex rate card:

The rate card adds three rules: Fast mode costs 2x the Standard credit rate, Ultrafast on GPT-6 Astra and GPT-6.1 Sol costs 6x, and Codex does not charge for cache writes. OpenAI's pricing page also says a typical GPT-5.6 Sol task may use 2 to 15 credits.

One gap matters for budgeting. None of the OpenAI help pages we checked for this guide list a dollar price per credit; they point to the rate card, which is expressed only in credits. Check the price shown at checkout before you buy a pack.

An API key is cheaper than credits whenever a credit costs you more than about 4 cents. That threshold comes from our own comparison of OpenAI's Codex rate card with its API price list: on every GPT-6 model we compared, the credit rate is exactly 25 credits for each $1 of API list price.

Put simply, 1 credit buys the same tokens as $0.04 of API usage at list price. If you pay more than $0.04 per credit, running extra Codex sessions on an API key costs less. If you pay less, credits win. Our OpenAI API pricing guide covers the full API price list, including batch and cache discounts.

Take a task that sends 200,000 input tokens (150,000 of them cached from earlier turns) and gets back 20,000 output tokens. Here is what it costs at Standard speed, using the rates above:

Run 100 of those GPT-6.1 Sol tasks past your plan limit and you need roughly 790 credits, the same as $31.50 at API list price. Switching the same work to Fast mode doubles the credit cost. The lesson is that model choice moves your bill far more than plan choice: Astra costs 100 times more per token than Luna.

The API route has a catch. OpenAI's pricing page says API-key Codex works only in the CLI, SDK and IDE extension, without cloud features such as GitHub code review or Slack.

For teams, Codex costs $20 to $125 per seat per month on ChatGPT Business, plus optional workspace credits. OpenAI's Business page lists a Standard seat at $25 per user per month (or $20 billed annually) and a Premium seat at $125 per month (or $100 annually) with 5x Standard usage and no five-hour limit. A workspace needs at least two seats.

Business seats each have their own limits. According to OpenAI's flexible pricing page, a user who exceeds them draws from a shared workspace pool if the owner has bought credits. Business credits last 12 months, aren't refundable except in narrow cases, and owners can turn on automatic recharge.

Enterprise and Edu work differently. The workspace buys one shared credit pool through its OpenAI account team, with no per-seat caps by default, and admins set spend controls by group. OpenAI's Codex pricing page says Enterprise and Edu plans with flexible pricing have no fixed rate limits, while those without flexible pricing get roughly Plus-level limits per seat. Enterprise agreements with USD billing use a separate dollar rate card.

Seat types have shifted this year. OpenAI added a Codex-only seat for Business and Enterprise on April 2, 2026, but its Business help page now lists the Codex seat as a legacy option for workspaces that had it before June 24, 2026. New teams should plan around Standard and Premium seats. Our guide to the ChatGPT Business plan covers the non-coding side of those seats.

Most individual developers should start on Plus and upgrade only when the five-hour window keeps stopping real work. Here is how the options line up by usage pattern:

OpenAI is pitching price as an edge over Anthropic. TechCrunch reported OpenAI's claim: "Compared with Claude Code, Codex delivers more coding capacity per dollar across paid tiers." That is the company's own claim, not an independent test, so compare against your actual workload. Our Claude Code vs Cursor comparison and Cursor pricing guide cover the main alternatives.

Sameer's take: He pays for Claude and ChatGPT. His view: AI subscriptions at about $20 a month are easily worth it for a professional.

Codex is included across ChatGPT plans, including Free and Go, according to OpenAI's help center. Free and Go users get GPT-6 Luna in the desktop app only, without Codex Cloud, so cloud tasks and code review need Plus or higher.

Codex comes with ChatGPT Plus at $20 a month, with no separate Codex fee. OpenAI estimates Plus users get about 15 to 160 local messages per five hours on GPT-6.1 Sol, depending on task size.

ChatGPT Pro now comes in three tiers: $100, $200 and $500 a month. Pro $200 includes more usage than Pro $100, and Pro $500 has the highest included usage and is the only Pro plan with Ultrafast mode. All Pro tiers include Codex, and Pro currently has no five-hour limit.

Codex limits run on a rolling five-hour window, and weekly limits may also apply. OpenAI Support does not reset limits, though some Plus and Pro referral promotions add a one-time banked reset.

Yes. Codex works in the CLI, SDK and IDE extension with an API key, billed at standard API rates, but cloud features such as GitHub code review and Slack are not included.

OpenAI Codex pricing comes down to three choices: which ChatGPT plan you pay for, which model you run, and what you do when the five-hour window runs out. Plus at $20 is the right starting point for most developers, Pro $100 is for people who code with Codex all day, and Business seats run $20 to $125 per user.

For overflow, remember the 4-cent rule from our rate card comparison: 1 credit buys the same tokens as $0.04 of API usage, so the cheaper path depends on what you pay per credit. OpenAI changes models and limits often, so we will keep this page updated as the rate card moves. For price changes across every major AI tool, see our monthly AI Pricing Index.

⭐ Get more AI Business Weekly in Google: add us as a preferred source on Google.

Written by Sameer Khan with AI assistance, fact-checked against the sources linked above.