OpenAI Codex pricing

There are two ways to pay for OpenAI Codex: through a ChatGPT plan or separately through API billing. If you sign in with ChatGPT, your plan covers usage up to its limits, with additional credits available on eligible accounts. If you sign in with an API key, you pay for the tokens Codex uses through your OpenAI Platform account. Your sign-in method also affects which features you can use.

How much do Codex plans cost, and what do you get?

Codex is included across ChatGPT plans, but the amount of usage and the features you get depend on your plan.

OpenAI Codex pricing for individual plans

For individuals, OpenAI offers a free plan and several paid plans, ranging from $8 to $500 per month. Go, Plus, and Pro are billed monthly, so there’s no annual commitment or annual billing discount.

If you’re starting with Free or Go, keep in mind that access is more limited. Both currently offer GPT-6 Luna at Standard speed in the desktop app, subject to rollout, and neither includes Codex Cloud.

OpenAI Codex pricing for business and enterprise plans

For businesses, OpenAI offers Standard and Premium seats, starting at $20 and $100 per user per month when billed annually. Monthly plans are also available, but they're at a higher rate. 

What are OpenAI Codex’s usage limits?

Codex and ChatGPT Work share usage, along with other eligible agentic features. Your local and cloud Codex tasks also draw from the same allowance, so moving a task to the cloud doesn’t give you a separate pool of usage.

Plus and Business Standard have five-hour usage windows. Pro and Business Premium currently don’t, although weekly limits may still apply.

How quickly you go through your allowance depends on the model you choose, the size of the task, and your reasoning and speed settings. For example, asking Codex to tackle a complicated refactor with a high reasoning setting can use much more of your allowance than a quick fix. Two requests don’t necessarily use the same amount just because each started with one prompt.

To stay on top of how much usage you have left, you can check Codex’s usage dashboard or enter /status in the CLI.

OpenAI has also announced a change starting in December 2026: Plus, Pro, and Business will share an allowance across ChatGPT and Codex, with Instant chat excluded. If you use both for work, that’s a change to keep an eye on.

What happens when you run out of included usage in Codex?

Hitting your limit doesn’t always mean you have to stop working or upgrade your plan. OpenAI allows users to buy ChatGPT credits to keep using Codex after their included usage runs out. Purchased personal-plan credits are valid for 12 months.

One detail worth keeping straight: these are ChatGPT credits, not API credits. Usage is charged according to the Codex credit rate card, and the price you pay for credits depends on your plan or agreement.

Can you use the API credits included with Pro and Business in Codex?

No. The subscription API credits offered with eligible Pro and Business plans can support other API work, but they won’t cover Codex usage through an API key or give you more usage when you’re signed in with ChatGPT.

How does OpenAI Codex API pricing work?

With API billing, you pay for what Codex uses at your chosen model’s API rates. Input tokens are the information the model reads; output tokens are what it generates. Those have different prices, and cached input is charged at a lower rate.

You can use an API key for local Codex work, but Codex Cloud requires ChatGPT sign-in. If cloud tasks are part of your workflow, make sure your plan includes that access.

What do OpenAI’s coding models cost?

The table below shows Standard API rates for short-context requests, in dollars per million tokens. These prices apply when you use an API key; they aren’t a way to calculate how much coding your ChatGPT subscription includes.

Other than subscription price, what else affects your Codex costs?

The model you choose is only part of the cost. A few other factors can change your bill or make your included usage run out faster:

• Caching: Reusing cached input costs less than processing it again as new input. On the API, writing that information to the cache has its own charge, so both reads and writes matter.
• Reasoning: Thinking tokens count as billable output, even though you don’t see them in the final answer. That means a short response can still involve more model work than its length suggests.
• Long context: For GPT-6.1 Sol API requests above 272,000 input tokens, the whole request is charged at 2x the input and cache rates and 1.5x the output rate. That long-context surcharge doesn’t apply to usage drawn from your subscription.
• Speed: Faster processing uses more of your budget. GPT-6.1 Sol API Fast mode costs 2x Standard, and Ultrafast costs 6x. When you’re signed in with ChatGPT, those modes use included allowance at 2.5x and 8x Standard; purchased credits are charged at 2x and 6x.
• Regional processing: If you choose a supported regional processing option for GPT-6.1 Sol, its API pricing includes a 10% premium.

Which Codex plan should you get started with?

If you’re just getting started with OpenAI Codex, you can start with the features you need, then see how often limits interrupt your work. 

• Free or Go: A place to try lightweight coding tasks, provided the available desktop access covers what you want to do.
• Plus: A starting point if you want broader Codex access and a chance to try it on your everyday work before paying for a larger plan.
• Pro: Worth considering if you regularly hit your limits. Compare the upgrade price with what you’re already spending on extra credits and how much the interruptions affect your work.
• Business: A fit when you need a shared workspace, team billing, and admin controls. You can mix Standard and Premium seats, so frequent users can get more capacity without upgrading everyone. You’ll need at least two seats.
• Enterprise: Worth discussing with sales if your organization needs controls such as SCIM, audit logs, or specific data retention and residency settings.
• API billing: An option for automated jobs or workloads that vary from week to week. Try a few jobs that reflect your actual work before setting a budget.

If some developers sign in with ChatGPT and others use API keys, remember that you’re managing two billing routes. ChatGPT workspace spending controls don’t cover API Platform billing, so you’ll need to track and control those costs separately.

How do you know if Codex is worth the cost?

You can judge whether Codex is worth the cost by connecting spend to work that ships: merged pull requests, resolved tickets, and released features. We call this cost per outcome, and it gives engineering leaders a way to decide whether to renew, expand, or cap their investment.

The invoice alone won’t tell you whether that investment is paying off. A high bill could mean developers are stuck in costly loops, or that they’re getting more valuable work done. Faros CEO Vitaly Gordon described one Faros engineer spending $20,000 on AI tokens in a month. Without knowing what that usage produced, nobody could tell whether the engineer deserved a promotion or a difficult conversation about spending.

Even developers on the same Codex plan can get very different results. Model choice, the context they give the agent, and how they divide up the work all play a part. Task difficulty matters, too: a high cost per outcome might reflect a difficult repository or unnecessary AI usage. Looking at the underlying pull requests and delivery outcomes helps you understand which. That’s why we encourage engineering leaders to move from token maxxing to outcome maxxing, starting with visibility into what each seat’s spend actually produced.

Track your OpenAI Codex costs with Faros

OpenAI Codex is one part of your AI coding bill. Faros brings costs across your AI coding tools into one view and connects that spend to engineering outcomes. Track costs by developer, team, and task, see what the investment produces, and use that evidence to guide model choices and spending decisions.

Reach out for a demo to learn more.

OpenAI Codex FAQs

Is OpenAI Codex free?

Yes, Codex offers limited access through ChatGPT Free, subject to rollout. Paid plans provide broader access and higher usage allowances. Free access doesn’t include Codex Cloud.

Is Codex included in ChatGPT Plus?

Yes. ChatGPT Plus includes Codex usage for $20 per month. Sign in with your ChatGPT account to use that allowance. If you connect an API key instead, usage is billed separately through the OpenAI API.

Does ChatGPT Pro include unlimited Codex usage?

No. ChatGPT Pro offers higher Codex allowances, but usage limits still apply. Pro plans currently cost $100, $200, or $500 per month. The amount of work each allowance supports depends on your model, task complexity, and settings.

Does Codex CLI cost extra?

Codex CLI doesn’t require a separate subscription. When you sign in with ChatGPT, it uses your plan’s included allowance and any applicable credits. When you authenticate with an API key, you pay separately at API rates.

Is a Codex subscription cheaper than API billing?

Neither option is always cheaper. A subscription provides included usage for a fixed monthly price, while API billing charges for the tokens your work consumes. Compare both using representative tasks, including any extra credits you would need beyond your subscription allowance.

Are Codex credits the same as OpenAI API credits?

No. ChatGPT usage credits extend eligible Codex usage after your included allowance runs out. API credits belong to the separate API billing system. The promotional API credits included with eligible Pro and Business subscriptions specifically exclude Codex usage through API keys.

When do Codex usage limits reset?

Reset times depend on your plan and the allowance you’ve exhausted. Check your Codex usage dashboard for the applicable reset time, or enter /status during a CLI session. Reaching one limit doesn’t necessarily mean every available model or feature is unavailable.

Why am I hitting Codex limits so quickly?

Codex usage depends on more than the number of prompts you send. Larger tasks, longer context, more demanding models, and higher reasoning or speed settings can consume your allowance faster. Local and cloud tasks also draw from the same allowance.

How can I reduce OpenAI Codex costs?

Start by testing lower-cost models on routine tasks and keeping assignments focused. Compare the cost of extra credits with a plan upgrade. Judge savings by the cost of completing usable work, including retries and human review, so cheaper requests don’t become more expensive outcomes.

How can a company track Codex spending across developers and teams?

Track subscription fees, additional ChatGPT credits, and API usage together. Then connect those costs to developers, teams, and completed work. Faros brings AI coding spend across tools into one view and connects it to engineering outcomes, helping leaders understand what their investment produces.