LLM API Cost Calculator
Turn a token count into money. Enter input and output tokens, or let the tool count them from pasted text, and see the cost for one call, a thousand calls and a month of calls across the models you care about, side by side.
Edit prices (per 1M tokens, USD)
Each row is editable. Blank or negative values are rejected. Prices are an editable input; the source is the provider's pricing page, which the user is expected to keep current.
Cost comparison, sorted by monthly total
| Model | Per call | Per 1,000 | Per day | Per month |
|---|
Count from text (optional)
Paste a prompt and an expected reply and the tool will fill the token fields with the estimated count for the selected estimator. This is an estimate using charCount / 4; for an exact count, use the Token Counter.
How to use the LLM API Cost Calculator
Enter the input and output token counts, set the call volume, and pick the currency. The table on the right sorts the included models by monthly cost, and the cheapest row is marked with a star. Click any number to copy it.
The prices are stored in an editable table at the top. Open the “Edit prices” panel to change any value. The defaults are placeholders, not market data, and the alert at the top of the page reminds you to check the provider pricing page before you budget.
What does the exchange rate field do?
It converts USD prices into INR for the table. The default is a placeholder; a hard-coded rate goes stale fast, so the field is editable. If you are billing in USD, set the rate to 1 and the table will read in dollars.
Are these the official prices?
No. The defaults are reasonable placeholders dated 2025-Q4. Every provider changes prices without notice, and there are volume tiers, prompt-cache discounts and other adjustments the calculator does not model. For a budget that a finance team will sign off on, get the prices from the provider's pricing page and paste them into the Edit Prices panel.
Why is the cheapest row not always the one I want?
Quality and capability vary a lot between models at similar prices. The cheapest row is the cheapest row, not the best row. The token cost is one input into a model decision, and the calculator deliberately stops at the cost so it does not pretend to know quality.