What Are AI API Credits? How Prepaid Billing Works

By the Zylo team · Updated June 8, 2026 · 5 min read

When you use a paid AI API, you need a way to pre-authorize spending before individual token charges accumulate. Most platforms — including Zylo — solve this with a credits system: you add a balance to your account, and each API call draws down that balance in proportion to the tokens consumed. Credits are not a subscription fee and they are not a separate currency with an opaque exchange rate; they map directly onto token costs. One dollar of credits at Zylo buys exactly one dollar worth of tokens at the published per-million-token rates for the model you call. There are no per-request fees, no seat fees, and no markup on usage. Credits simply act as a prepaid wallet that lets you control spending, set usage limits, and avoid surprise invoices at the end of the month. Understanding how credits are funded and how they are spent is the first step toward accurate cost planning for any AI-powered product.

The 25 percent platform fee at top-up

Zylo charges a flat 25 percent platform fee — but it is critical to understand exactly when it applies: only when you add credits to your account, never on usage itself. If you top up $100, the platform fee is applied at that moment, and from then on every token call is billed at the exact published per-million-token base rate with zero further markup. This model is intentionally transparent: you see the fee once, at the moment you fund your account, and then you can plan usage costs using the public rate card without any hidden multipliers. It contrasts with platforms that silently inflate per-token prices to cover infrastructure costs invisibly. For accurate budgeting, account for the 25 percent fee in your total funding need rather than in individual token estimates. The Zylo cost calculator lets you enter a budget and see how many tokens or words that translates to at each model’s base rate.

The free Basic plan and what it includes

Zylo’s free Basic plan does not include credits. You do not need a credit card to sign up, and no billing information is required to obtain an API key. The Basic plan provides approximately 200,000 tokens and 7,200 requests per day, subject to a 10-request-per-minute rate limit and a 200,000-input-character cap per request. These allowances cover Basic-tier lightweight models only; premium models such as Claude Opus 4.8, GPT-5.5, or Gemini 3.1 Pro are not accessible on the Basic plan and require a paid plan with credits. The distinction matters for honest evaluation: the free aspect of Zylo covers the API key, the account creation, and a meaningful daily allowance on capable lightweight models — it does not mean unlimited or free access to flagship models. Developers exploring the API or running low-volume automations can often operate entirely within the Basic plan limits. For everything beyond that, the Zylo pricing page lists the Go, Pro, Mega, and Enterprise plans along with the credits each includes.

Credits versus markups and subscriptions

It helps to place credits beside the two other ways platforms charge for AI, because the differences decide your real cost at scale. A per-token markup adds a percentage on top of every unit you consume, so it scales directly with usage and quietly raises your effective rate exactly as your product grows. A flat subscription charges a fixed monthly amount regardless of how much you call, which can either waste money on light months or cap your throughput on heavy ones. Credits sit apart from both: they are simply prepaid value spent at the base rate, with Zylo’s only platform fee charged once at top-up rather than on consumption. The practical consequence is predictability — your per-token cost is the published rate and does not drift upward with volume, so a forecast you build at a thousand calls a day still holds at a million. When you compare providers, look past the headline rate to whether usage itself is marked up, because that single difference often matters more than the sticker price.

How to think about credit budgeting

Practical credit budgeting starts by estimating the token volume a workload will generate, then converting that to a dollar figure using the model’s published rates, and finally accounting for the 25 percent top-up fee when funding the account. For example, if your application is projected to consume 10 million output tokens per month on Gemini 2.5 Flash Lite ($0.40 per million output, $0.10 per million input as of June 2026), the pure token cost for output is $4. Add proportional input tokens — say another $1 — and your monthly token spend is approximately $5. Spreading top-ups over time rather than adding a large lump sum at once lets you calibrate spending as actual usage patterns emerge. The guide on AI API costs provides a broader framework for comparing model tiers, and the free API key walkthrough covers exactly what you can test before committing any budget.