How Pricing Works

Pricing

Part of: Cost & Token Budgeting

You send a model a one-line question and a dense three-page answer comes back. When the bill arrives, those three pages cost far more than your one line, and not only because they are longer. Once you see how LLM pricing works, the invoice stops surprising you. What it is An LLM API doesn't charge per request or per minute. It charges per token, the small chunks of text the model reads and writes. Every call has two token counts: the input tokens (your prompt, plus any history) and the output tokens (the model's reply). They are priced separately, and that split drives the cost. Prices are quoted per million tokens. You'll see a model listed like "$3 / 1M input, $15 / 1M output." The word "million" can look alarming, but most single calls use a few hundred to a few thousand tokens, so a call costs a fraction of a cent. The per-million number is a compact way to write a very small per-token price. How it works Two facts drive every bill: 1. Input and output are billed at different rates. The same provider that charges $3 per million input tokens might charge $15 per million output tokens. Reading your prompt is cheap; generating new text is expensive. 2. Output is usually several ti

Challenge: The Billing Meter