How to Buy OpenAI API Credits (and Skip the Hassle)

Buying OpenAI API credits is a five-minute job once you know where the button is: open the billing overview in your account settings, add a payment method, and confirm an initial credit purchase. OpenAI uses prepaid billing — you pre-buy usage, and API calls draw it down first. This guide walks the exact steps, the $5 auto-recharge minimum, and the two things that surprise people later: credits expire after one year and are non-refundable. Then the honest part — if you also call Claude, Gemini or open-source models, buying separate credit pools per provider is the hassle. One gateway balance spread across every provider skips it.

The short answer

Open the billing overview in your OpenAI account settings, add a payment method, and confirm an initial credit purchase. OpenAI runs on prepaid billing: you buy usage up front, and API calls draw it down first. During setup, Auto recharge is on by default — you set a recharge amount, a balance threshold that triggers it, and an optional monthly cap. When your credits run out, you return to the same billing portal and hit Add to balance to buy more. Two facts to file away now: the minimum auto-recharge amount is $5, and purchased credits expire after one year. Everything below is the detail.

Buy OpenAI API credits: the exact steps

The whole flow lives on the billing page — you do not need a support ticket or a sales call. Per OpenAI's setup article, the sequence is:

1. Open the billing overview. In your OpenAI account settings, go to the billing section — this is where balance, payment methods and invoices all live.
2. Add a payment method. If you have not already, add a card. Prepaid credits are charged to it.
3. Configure Auto recharge. It is enabled by default. Set your recharge amount, the balance threshold that triggers a top-up, and (optionally) a monthly recharge limit — or disable it entirely if you want to control every purchase by hand.
4. Confirm the initial credit purchase. This is the actual buy: you pick an amount and confirm, and the credits land on your balance.

That is it — you now have credits and the API will start drawing them down on your next call. Later top-ups use the Add to balance button in the same portal (covered below).

How OpenAI prepaid billing actually works

Prepaid credits are pre-purchased usage that gets applied to your monthly invoice — API usage is deducted from your credits first. The mechanics, straight from OpenAI's prepaid-billing explainer:

• You pre-buy a credit balance.
• API usage is deducted from prepaid credits first.
• If your usage exceeds the credits you purchased, you are billed for the additional amount on your invoice.

So prepaid credits are not a hard spend cap by themselves — they are a balance you draw down, with overage billed after. If you want a true ceiling, that is what the auto-recharge monthly limit and your usage limits are for.

Auto-recharge, minimums and limits

Auto recharge is on by default and has three knobs — set them deliberately, because the defaults keep buying credits for you. From OpenAI's setup article:

The $5 minimum is the documented floor for the auto-recharge amount; the maximum is defined by your Trust Tier. OpenAI does not publish a separate documented minimum for the one-time initial purchase distinct from that $5 auto-recharge floor — so treat $5 as the practical minimum and confirm against your own billing page. If you would rather not have credits bought automatically, disable auto recharge and rely on manual top-ups.

Why credits expire — and can't be brought back

Purchased prepaid credits expire one year (12 months) from the date of purchase, are non-refundable, and cannot be extended. This is the part that catches teams who buy a big block up front "to be safe." Per OpenAI:

• Credits expire after 1 year from purchase and are non-refundable; unused credits do not roll over past the expiry date. (source)
• OpenAI is unable to extend the expiration of, or reinstate, expired credit grants or purchased credits. (source)
• Enterprise and Education plans are the exception: credit allocation and expiration follow the terms in the customer's Order Form, which may differ from the standard 1-year API prepaid expiry.

The practical takeaway: buy in amounts you will actually consume within twelve months. A large speculative top-up that outlives its year is money you cannot get back. If your usage is spiky or you are not sure how much you will need, smaller top-ups (or auto-recharge) beat one giant prepayment.

Topping up after you run out

When your credits are consumed, go back to the billing portal and use "Add to balance" to purchase more. There is nothing more to it — same page, same payment method, same $5 minimum on the purchase. If you left auto recharge enabled, this happens for you the moment your balance crosses the threshold you set; if you disabled it, Add to balance is the manual path. Either way, remember the expiry clock starts fresh on each purchase, so a new top-up gives you a new twelve-month window on that amount.

Per-provider credits vs one gateway balance

Buying OpenAI credits is easy; buying — and tracking — a separate credit pool for every model provider is the actual hassle. The moment you also call Claude, Gemini or an open-source model, you are maintaining a payment method, a balance, a threshold and a 12-month expiry clock per vendor. A gateway collapses that into one balance you spend across all of them through a single OpenAI-compatible API. Here is the honest comparison:

To be clear about what is what: OpenAI's prepaid credits are the vendor's own billing product, and if OpenAI is the only provider you will ever call, buying credits directly is perfectly fine. The gateway wins when your stack is plural — one balance, one key, one base URL, and any provider you add later is already reachable. If cost is the driver, our cheapest-LLM breakdown compares per-token prices across providers on that single balance, and the gateway primer explains the routing underneath.

Switching is a two-line change — point your OpenAI SDK at the gateway base URL and keep your existing code:

from openai import OpenAI

client = OpenAI(
    base_url="https://www.datallmlab.com/v1",   # one balance, every provider
    api_key="YOUR_DATALLMLAB_KEY",
)

resp = client.chat.completions.create(
    model="openai/gpt-oss-120b",               # or claude-sonnet-5, glm-5.2, ...
    messages=[{"role": "user", "content": "Hello"}],
)

The gateway is OpenAI-compatible, so the same base URL also exposes a /v1/embeddings endpoint for embeddings workloads — you keep the OpenAI request shape and one balance covers it.

FAQ