Free tool

AI Usage Cost Calculator

Estimate what AI usage will cost, and compare options side by side. You enter current prices from each provider's official pricing page, because prices change often.

1. Your usage

Don't know your token counts? Convert words to tokens

English text averages roughly 1.3 tokens per word; code and other languages often use more.

2. Prices to compare US dollars per 1 million tokens

Option 1
Option 2 (optional)
Option 3 (optional)

Estimated cost

OptionPer requestPer dayPer monthPer year

Copyable summary

An estimate using the prices you entered. Real bills can differ: taxes, minimum charges, discounts for cached or batched requests, and extra fees for tools, web search, images or audio aren't included. Subscription plans (a fixed monthly price) work differently. JymboAI doesn't supply or check prices; always use the provider's official pricing page.

Worked example · fictional rates

Is the cheaper model really cheaper?

These rates are invented for teaching. They are not the prices of any real model or provider. Use current official prices for your own estimates.

Imagine a small helper that drafts replies to customer emails. Each request sends about 1,500 input tokens and gets back about 500 output tokens, 100 times a day, for 30 days a month.

Fictional optionInput rateOutput ratePer requestPer month
Model Alpha (fictional)$2.00 / 1M$8.00 / 1M$0.0070$21.00
Model Beta (fictional)$0.40 / 1M$1.60 / 1M$0.0014$4.20

The math for one Alpha request: 1,500 × $2.00 ÷ 1,000,000 = $0.0030 for input, plus 500 × $8.00 ÷ 1,000,000 = $0.0040 for output, so $0.0070 in total. Multiply by 100 requests and 30 days: $21.00 a month.

Now count finished work

Suppose Beta's drafts often need redoing before they're usable. What matters is the cost of an acceptable reply:

Beta attempts per usable replyCost per usable replyPer monthCompared with Alpha ($21.00, assuming one attempt)
1$0.0014$4.20Much cheaper
3$0.0042$12.60Still cheaper
5$0.0070$21.00Break-even
6$0.0084$25.20More expensive

This leaves out the time a person spends checking and correcting each draft, which often matters more than the model bill. Test both options on your own real examples before deciding.