Credits

AI usage is hard to price with one flat fee. A short question to a small model costs almost nothing, while a long document sent to a large model, a batch of images or a few seconds of video costs far more. Credits put all of that into a single unit. Instead of juggling prices per million tokens, per image and per second, you hold one balance, and every action takes its share. Prepaid units are familiar from phone cards, arcade tokens and transit passes: you pay once up front, and each use deducts a measured amount, which keeps spending visible at every step instead of arriving as one large bill later. The unit hides the arithmetic of exchange rates and provider invoices, leaving one familiar yardstick.

A prepaid balance has a second benefit: it is a ceiling. You cannot spend more than you loaded, so a buggy loop or a leaked key cannot produce a surprise bill. The price of a call is usually computed from what the provider charges: tokens in and tokens out for text, a fixed amount per image, a rate per second for video. Many services reserve the most a call could cost before it runs, then settle the real amount afterward and give back the difference. It helps to separate balance, spend and reservation: the balance is what is available, the spend is what was charged, and a reservation is a temporary hold while the call runs.

When you compare offers, check a few things. Is the unit tied to a currency, so that you can compare prices? Do credits expire, and can they be refunded? Are taxes included? And are there separate pools, for example one for a subscription and another for programmatic use? A clear service answers each of these in writing. It also pays to look at what the unit measures. A fair scheme ties credits to real cost, so a heavier request takes more and a lighter one less, rather than hiding a flat charge behind a vague allowance. Rounding rules matter too: tiny fractions per call add up across thousands of requests.

On NeverNo, 1 credit is $0.01. The API is billed at the real cost of the model you call, times 1.5 for chat models, open and closed alike (Claude, GPT, Gemini, Grok), and times 2.5 for images, with no subscription required. It draws on its own API balance: a plan's included credits and credits bought for the app are never spent by the API, and the API never spends them. You top up in Settings, from $10 to $2,000, taxes included and charged in your local currency. The first time you create an API key, your account also receives a starter credit of $1, once, usable on open models and images. Cost grows with the work done, so a brief chat with a small open model weighs far less than a long video or a large image.

Before a call, the most it can cost is reserved from your balance. When it finishes, you pay the real amount and the rest comes back; a call that fails on the provider's side is refunded. If the balance is too low, the API answers 402 with insufficient_credits. The price of every model in credits is listed in the pricing section of the API documentation and by the public models endpoint, which is where the real figures live. Because the multipliers and per-model prices are published, you can estimate a job in advance: multiply the tokens you expect by the listed rate, or take the per-image amount, and compare it with your balance.