LLM Token Cost Calculator
Model what an AI feature actually costs — per request and per month, including prompt caching and batch discounts.
Runs entirely in your browser — nothing you enter leaves this page
Prices last reviewed July 2026 — always editable below. Claude Sonnet 5 has introductory pricing ($2 / $10) through August 2026.
Estimate tokens from pasted text
Per request
Per month
How the math works
LLM APIs bill per token, with separate rates for input (your prompt plus
any context) and output (the model's response). Monthly cost is simply
(input tokens × input price + output tokens × output price) ×
requests. The two levers that change the bill most in practice are
prompt caching — repeated prompt prefixes are re-read at
roughly a tenth of the input price — and batch processing,
which trades latency for a 50% discount.
What's a realistic cache hit rate?
Chat apps and agents that resend a large system prompt every turn routinely see 70–90% of input tokens served from cache. One-shot classification with unique inputs sees close to 0%. We walk through the mechanics, TTLs, and break-even math in our guide to prompt caching.
Where do the preset prices come from?
Published list prices, last reviewed July 2026. Providers change pricing — every field above is editable, so plug in current rates if a preset has drifted. Token estimates from pasted text use the ~4-characters-per-token rule of thumb; for background on what a token even is, start with What is an LLM? and What is a transformer?.