Home / Free tools / Prompt cost analyzer
Free tool

Prompt cost analyzer

Paste a prompt to see roughly how many tokens it uses, what it costs per call, and how much you would save by moving the heavy parts out of the chat. No sign-up.

Token estimate uses roughly four characters per token. Real counts vary by model tokenizer.
How often this prompt runs, for the yearly figures.
How much of it is raw data that could be processed first.
Paste a prompt to analyze.
Estimate
Estimated tokens
0
Cost per call
$0
at $3.00 per million input tokens
Cost per year
$0
Saving per year if offloaded
$0

Estimate only. Token counts use a four-characters-per-token approximation and a rate of $3.00 per million input tokens. The 96.5 percent figure is a measured result on a data-heavy task, not a guarantee for every prompt.

How prompt cost is calculated

Large language models charge per token, and a token is roughly four characters of English text. The more you paste into a prompt, the more tokens it carries and the more each call costs. This tool counts the characters in your prompt, estimates the tokens, and prices them at three dollars per million input tokens. Multiply by how often you send it and you get the yearly cost.

The biggest waste is raw data: pasting a whole log, spreadsheet, or web page into the prompt so the model can reason over it. If that work runs through an execution layer first, only the small answer comes back, and the token count on data-heavy prompts drops sharply.

Frequently asked questions

How accurate is the token estimate?

It is a close approximation. Most models tokenize English at roughly four characters per token, so the estimate is usually within a small margin. Code, other languages, and unusual formatting can shift the real count. For exact numbers, check your model provider tokenizer.

How do I lower a prompt cost?

Stop pasting raw data into the prompt. Send an instruction and let the heavy work happen outside the chat, then read back only the result. On data-heavy prompts this is where the largest saving comes from, up to 96.5 percent on a measured log task.

Does this work for ChatGPT and Gemini?

Yes. The token and cost model is similar across major providers, and UniversalBench connects through one MCP link that works with Claude, ChatGPT, Gemini, and any MCP-compatible AI.

Cut the cost of heavy prompts

Move the raw data out of the chat. Connect one URL and pay for far fewer tokens. First 1,000 calls free.

Get your API key →