Codex CLI Pricing: ChatGPT Plans vs API Key Billing

You have just run codex, a sign-in prompt has appeared, and you are not sure which choice sends the bill where.

Two ways to pay for Codex CLI

To use a plan, open your project directory, run codex and sign in with your ChatGPT account when prompted. Codex is included in ChatGPT Free, Go, Plus, Pro, Business, Edu and Enterprise plans.

The other route is an OpenAI API key. Codex then runs in the CLI, SDK or IDE extension, and usage is charged at API prices. Model availability follows the API models available to your key.

The API key route has no cloud-based features such as GitHub code review or Slack. That makes it a clearer fit for automation in shared environments such as CI, where per-token billing may be easier to allocate.

Codex CLI billing routes, checked 3 November 2026

Source: OpenAI Codex pricing pages and documentation, read 3 November 2026. Both official pages showed different Go prices, so both are reported instead of choosing one.

How usage limits work

ChatGPT Work and Codex share usage, credits and usage limits. Local messages and cloud chats also share your plan's allowance, and weekly limits may apply.

How many messages you can send depends on the model, the size and complexity of the task, and whether it runs locally or in the cloud. Pro plans currently have no five-hour limit.

If you hit a limit during an active turn, the agent can continue that turn, subject to fair use limits. You can check the Codex usage dashboard for current limits. In an active CLI session, run /status to see what remains.

Auto-review safety checks are free when you are signed in with a ChatGPT account and do not count against the plan allowance.

Credits pay for more than your plan includes

Credits are the unit used for eligible usage on credit-based plans. Plus and Pro users can buy more credits after reaching a usage limit instead of upgrading immediately. Purchase prices and discounts depend on the plan or agreement.

The rate card below is for purchased credits at Standard speed. Each figure is credits per million tokens.

Codex purchased-credit rate card, checked 3 November 2026

Source: OpenAI Codex pricing documentation, read 3 November 2026. Fast mode uses twice the Standard credit rate where available, while Ultrafast uses six times the Standard rate. Ultrafast mode is available on Pro $500, checked 3 November 2026.

Credit billing has no separate cache-write charge. API-key usage follows the separate API price card, so these credit figures should not be used to estimate an API bill.

Our example: ten tasks in one week

Our example is illustrative and uses OpenAI's published typical range. A GPT-5.6 Sol task may use 2 to 15 credits. Our calculation: ten similar tasks would use 20 to 150 credits.

Our calculation for one task, using the Standard rate card checked 3 November 2026: assume one million input tokens, one million cached input tokens and one million output tokens. GPT-6.1 Sol would use 302.5 credits. GPT-5.6 Sol would use 610 credits.

For this token mix, GPT-6.1 Sol uses fewer purchased credits than GPT-5.6 Sol, so the model choice changes how far the same credit balance goes.

The actual total can move outside that example because task size and complexity differ. There is no dollar total here because credit purchase prices depend on the plan or agreement.

The API key route: when per-token billing makes sense

An API key is useful when a team needs Codex in CI or another shared automation environment. Each token is billed through the API account, without drawing from a person's ChatGPT allowance.

The trade-off is direct: API-key sessions do not include Codex Cloud features such as GitHub code review or Slack. They also follow the models available to that API key.

OpenAI says API token prices are separate from subscription usage and should not be used to estimate included tasks. GPT-5.5 retired from Codex with ChatGPT sign-in on 14 October 2026, but API and API-key sessions were not affected.

Which billing route should you pick?

Choose the route that matches where the work runs and who should receive the bill.

A practical route by working pattern

For routine work, switching to a lower-cost model can help the allowance last longer. Check /status before a large task so the route and remaining allowance are clear.

The sign-in choice determines the billing route.

Where Xtrusio fits

Codex helps teams change code. Xtrusio helps them inspect AI answers and plan a response. Its free visibility check shows when ChatGPT and Gemini mention a business, which competitors appear instead and what to work on next.

Run the free AI visibility check when the question moves from coding cost to how a brand appears in AI answers.

Sources

• OpenAI Codex pricing documentation, read 3 November 2026
• ChatGPT Codex pricing page, read 3 November 2026
• OpenAI Help Center: using Codex with a ChatGPT plan, read 3 November 2026
• OpenAI Codex CLI documentation, read 3 November 2026