Difference between prepaid and pay as you go API Pricing

What is the difference between prepaid and pay-as-you-go pricing on API?
For example, in the pay-per-use method, we pay $30 for each million GPT4 output tokens. Does the price different in prepaid payment method?

Pricing | OpenAI vs Pricing - OpenAI API

I cannot tell which is which and if it expires. Hope someone can advice


When you use the API, all payment methods are pay-as-you-go and prepaid.

Both apply at the same time, meaning they are not mutually exclusive.

Let’s go over this with an example from OP and another one.

For GPT-4 (which probably refers to GPT-4 Turbo), the cost is $30 per 1M tokens.

GPT-4o and GPT-4-mini are different models with separate pricing.

Right now, only the prepaid method is accepted, but even with that, the price stays the same.
And prepaid credits stay valid for one year after purchase.

But you might wonder whether older credits get used first—I find this a bit confusing myself.

So, if you’re worried about that, you can either wait until your balance is low before recharging. Then, if your balance goes negative and auto-recharge is off, you will need to manually recharge to continue using the service. (I’m not entirely sure, but I believe that’s how it works.)

This might appear a bit off-topic, but:

Even if you turn on auto-recharge, you can still set a monthly spending limit using Budget Alert and Budget Limit.

When your spending hits the Budget Alert amount, you’ll get an email notification.

And when you reach the Budget Limit, your usage for that month will be paused.

But please keep in mind, there’s a delay in detecting when you go over the Budget Limit, so you might still be charged beyond it, and in some cases, the excess could be substantial depending on usage speed.

This feature helps keep spending under control, but it’s not a hard limit.

If you want strict control over your spending, perhaps the best approach is to turn off auto-recharge and only use the amount you’ve prepaid.

I hope this helps!:slightly_smiling_face:

1 Like

In the past, an API account was “pay as you go”, a phrase perhaps just to contrast your usage-based billing on the API from that of a ChatGPT subscription that is recurrent. Here’s that phrase showing even with monthly post-payment:

With virtually everyone being forced off of post-pay and needing to make full prepayment satisfactory for any usage job, and needing to meet higher levels of handing over cash in the past to activate higher “tier” limits and and have a chance at being offered newest models, the API service can barely be called “pay as you go” now, despite that text still in the user interface.


It is “Now pay minimum amounts of dollars, buying credits”. Credits which can then be used to fund individual API calls - until the credits self-expire in a year, or any OpenAI-manufactured account termination reason.

(FYI: You need to do your prepay credit check on OpenAI’s trustworthiness - instead of the other way around for a typical netX business invoice. You’re not truly loaning OpenAI money for a year though; they instantly convert cash into promises and disclaimers. The money is not coming back out to you, service or no service. )

OpenAI currently has two API pricing pages that both serve similar purposes answering about the same models and pricing.

1 Like