Question 1

What is a token in AI?

Accepted Answer

A token is a small piece of text that an AI model reads and bills for — the basic unit behind every token calculator, whether you call it a GPT token calculator, an AI token estimator, or an LLM cost calculator. In English, one token is roughly 4 characters, or about three quarters of a word. For example, “Hello world” is 2 tokens, while a longer word like “unbelievable” may split into 2 or 3. Providers charge per token, so more tokens means a higher cost.

Question 2

Why do output tokens cost more than input tokens?

Accepted Answer

Generating a reply takes more computing work than reading your prompt, so providers charge more for output tokens — usually several times the input rate. For example, Claude Sonnet 5 costs $2 per 1M input tokens but $10 per 1M output tokens. That means 1,000 input tokens cost $0.002, while 1,000 output tokens cost $0.01. A long reply can end up costing more than a long prompt, which is why the Reply size setting in this token cost calculator matters so much.

Question 3

Is my prompt sent to a server?

Accepted Answer

No — your prompt never leaves your device. Token counting and cost math run entirely inside your browser, and PromptCost has no sign-in. This page uses Google Analytics to see how many people visit, but that only counts page views; your prompt content is never read, stored, or sent anywhere. Prompts you choose to save are kept only in this browser.

Question 4

Why are some prices marked third-party?

Accepted Answer

Official means the price comes straight from the provider’s own pricing page. Third-party means it was taken from a price-tracking source instead. Third-party API pricing is usually close to accurate, but it can lag behind real changes, so always check the provider’s own pricing page before setting a budget.

Question 5

How do I reduce my AI API costs?

Accepted Answer

Four easy ways to cut LLM API costs: 1) Trim your prompt and remove text the model doesn’t need — fewer input tokens, lower cost. 2) Set a lower reply size, since output tokens cost more per token. 3) Use a smaller, cheaper model for simple jobs like labeling or short answers. 4) Ask your provider about discounts for cached (repeated) input and batch jobs. Use the tokens calculator and comparison table above to see how much each change saves.

What will this prompt actually cost?