Free · in your browser · any OS

AI cost and token calculator

Paste your text or enter a token count, say how often you send it, and compare the monthly cost of cloud AI models with running a model on your own PC. Every price is an editable cell.

FreeNo signupWindows · macOS · Linux

1. Your usage

Paste a typical request (optional; the text stays in this page)

Input tokens per request (type it, or use the estimate above)

Expected output tokens per request

Requests per day

Days per month you use it

This is an estimate. Tokens are counted as characters ÷ 4, a common rule of thumb for English and other Latin-script text. Other scripts (for example Arabic, Chinese, Hebrew, Hindi, Japanese, Korean, Thai) and code often use noticeably more tokens for the same text, so type a real count if you have one. Each model family also counts tokens a little differently. In a chat, earlier messages are sent again with each new one, so the input per request grows during a conversation.

2. Optional discounts (you type them)

Cached-input discount, % off the input price (0 = off)

Share of your input that is cached, %

Batch discount, % off input and output (0 = off)

Some providers offer cheaper cached input or batch processing. Whether and how much differs by provider and changes. These boxes are your own assumptions, not facts: take the percent from the provider's pricing page.

3. Cloud prices, USD per 1M tokens (edit any cell, add your own row)

Model	Vendor	Input $	Output $

4. Local model on your own PC

Power draw while it works, watts

Hours per day it is actually running

Electricity price per kWh (your currency, as in your bill)

Price of a GPU you would buy for this (same currency; 0 = already own)

Compare the break-even against

The local row costs $0 per token; its only running cost here is electricity. The electricity price and GPU price are in whatever currency you type, and the cloud prices are in USD, so use the same currency for both or convert first.

Result: monthly cost, cheapest first

#	Option	Per request	Per month	Per year

Price list last checked . Quality, speed and limits differ between models, and this table only compares price: a cheaper row is not necessarily able to do the same job.

How it works

Cost per request = (input tokens × input price + output tokens × output price) ÷ 1,000,000, then your cached-input and batch percentages are applied if you typed any. Cached input: only the share you mark as cached gets the cached discount. Monthly = per request × requests per day × days per month. The local row is electricity only: watts ÷ 1000 × hours per day × days × price per kWh. Break-even = GPU price ÷ (monthly cloud cost of the row you chose − monthly electricity); if the cloud row is not more expensive than electricity there is no break-even. The break-even ignores resale value, the rest of the computer, your time, and any difference in answer quality.