Why Your OpenAI API Bill Is Separate from ChatGPT

Your OpenAI API bill is separate from ChatGPT because you are buying two different products that happen to share a company and a login. ChatGPT plans, $20 Plus through $200 Pro, are flat subscriptions to the ChatGPT product, billed where you manage ChatGPT. The API is a developer platform that sells tokens by the million, billed where you manage keys. Money paid to one never appears in the other, in either direction. The one documented bridge between them is Codex, which is included in ChatGPT plans and runs programmatically.

This is the most common billing confusion in every automation community we read, so consider this page the PSA version.

The error that brings everyone here

The canonical sequence: you pay for Plus, you create an API key for n8n or a Python script, you run the first workflow, and OpenAI answers with this:

{
  "error": {
    "message": "You exceeded your current quota, please check your plan and billing details.",
    "type": "insufficient_quota",
    "code": "insufficient_quota"
  }
}

Nothing is broken. The key is drawing on your API platform balance, which is empty, while your $20 sits in the other product. The key has no idea your Plus subscription exists, and no setting connects them.

Two products, two bills

	ChatGPT (subscription)	OpenAI API (platform)
What you buy	Flat plan with usage windows	Tokens, metered per million
Where billing lives	Your ChatGPT account settings	The developer platform’s billing page
Credential	Your login session	API keys you create
Bill shape	$20-200/month, fixed	Scales with usage, no ceiling
Includes Codex	Yes, every plan	Codex can also run on a key
Affects the other’s balance	Never	Never

The practical consequences, stated flatly: a Plus or Pro plan never discounts a token. API credits never raise ChatGPT limits. Canceling either changes nothing about the other. If you use both, you will receive two charges from OpenAI every month, and that is the system working as designed.

Current per-token rates, model by model, live in OpenAI API pricing explained.

Why OpenAI keeps them separate

The products sell different economics to different buyers. The API sells unbounded, metered capacity to developers who need streaming, the full model catalog, embeddings, and fine-tunes, and who pay for exactly what they consume. ChatGPT plans sell predictable flat pricing to people, with rolling usage windows instead of a meter as the constraint. Merging the billing would force one pricing philosophy onto the other product, so the wall stays.

The wall is also why “I already pay for ChatGPT, why am I paying twice?” has no settings-page answer. The inventory of what each product actually contains is in does ChatGPT Plus include API access?

Check both ledgers in two minutes

If you are unsure what you actually pay OpenAI, look in both places, because each page shows only its own product. Your ChatGPT plan and renewal date live in the account settings at chatgpt.com. API balance, per-key usage, and invoices live on the developer platform’s billing and usage pages.

That two-minute audit regularly surfaces two findings: a subscription someone forgot, and an API key a workflow is quietly draining. And if the platform page shows an empty balance while your automations throw insufficient_quota, you have located the wall this article describes, not a bug.

The one bridge: Codex is in the plans

One thing crosses the wall, and OpenAI built it. Codex, OpenAI’s coding agent, is included in every ChatGPT plan, and it is programmatic by design: codex exec is the CLI’s documented non-interactive mode for scripts and CI, and device-code sign-in is documented for headless machines. Work running through Codex under your ChatGPT sign-in bills to the flat plan, not to a key.

As planning estimates, a $20 Plus plan absorbs roughly $700 of API-equivalent Codex work per month, Pro 5x at $100 roughly $3,500, and Pro 20x at $200 roughly $14,000; estimates, never guarantees. Exposed as an OpenAI-compatible endpoint, that capacity slots into the same tools that produced the insufficient_quota error, which is precisely what using your ChatGPT subscription as an API covers end to end, and what the n8n walkthrough shows for the workflow case.

The split to remember: keep streaming, embeddings, and full-catalog work on the metered key, and put bulk programmatic volume on the plan.

Two products, two bills, one bridge. If your metered bill has grown past a couple hundred dollars a month, the calculator shows what the bridged version of it costs.