Free calculator
LLM API cost calculator
What your AI feature costs per month on each model, at the providers' standard API prices. Enter your traffic once and compare Claude, OpenAI and Gemini side by side.
| Model | Input / output per 1M | Per month |
|---|
Standard tier list prices from anthropic.com, developers.openai.com and ai.google.dev, checked 2026-10-11; 30-day month. Gemini prices are for prompts up to 200k tokens and cache storage per hour is not included. Batch APIs are about half price; cache writes and tool fees are not included.
Where LLM bills usually come from
Input tokens, mostly. A chat feature that resends the whole conversation on every turn pays for the first message again on turn two, three and four, and by turn ten the input is many times the size of the answer. Put that conversation behind a long system prompt and a set of tool definitions and the input side of the bill grows fast.
The parts that repeat on every call, the system prompt, the tool definitions, a reference document, can usually be served from the provider's prompt cache, where cached input costs roughly a tenth of the normal price on Claude and OpenAI. Set the cache share above to see what that does to your number. Work that does not need an answer within seconds, nightly summaries or tagging a backlog, can go through the batch APIs at about half price.
Switching to a cheaper model is the last lever, not the first. A smaller model that needs a retry or two can cost more per finished task than a bigger one that gets it right once, so measure on real requests before you switch.
If your AI bill grew faster than your users, send us the usage page.