LLM API Pricing Calculator
Compare the monthly dollar cost of calling different LLM APIs for your own workload -- fresh input, cached input, and output tokens priced separately, using the rates you paste in from each provider's own pricing page.
Bring your own rates. This tool does not fetch or assert any provider's current price. Paste the input, cached-input, and output rates from each vendor's own pricing page, current today, and verify before you commit spend. The preloaded numbers below are illustrative examples, not current prices.
Why cached tokens usually dominate an agentic workload's bill
This calculator vs. a blended per-token estimate -- how far off one flat $/M rate can land.
Already have a usage log? The LLM Usage Log Analyzer shows where your actual bill went.
Comparing a subscription AI coding plan instead of raw API usage? See the AI Coding Plan vs. API Cost Calculator.
Paying for compute by the hour as well? The AWS commitment break-even calculator finds how much to commit to from your own usage curve.
1. Monthly token workload
The tokens you send and receive each month -- the same for every provider, since the workload doesn't change, only the rate it's billed at.
2. Provider rates ($ per million tokens)
Name each provider and paste in its currently quoted input, cached-input, and output rates from that provider's own pricing page. Leave a row's name blank to drop it from the comparison. The values below are example placeholders -- replace them with the current rate from each provider's pricing page.
Provider A
example -- replace with the current rate
example -- replace with the current rate
example -- replace with the current rate
Provider B
example -- replace with the current rate
example -- replace with the current rate
example -- replace with the current rate
Provider C
example -- replace with the current rate
example -- replace with the current rate
example -- replace with the current rate
Custom provider
Monthly comparison
Usage is optional, stays in this browser, and counts only calculations and copied summaries. No inputs, identifiers, cookies, or analytics service are used. Uncheck to turn it off. Do Not Track and Global Privacy Control disable counting automatically.
Common questions
Why does this calculator ask for cached input tokens separately from fresh input?
Because modern LLM APIs bill them at very different rates. A cached input token — one the provider has already processed recently, such as a repeated system prompt or the prior turns of an agentic loop — is commonly billed at a fifth to a tenth of the fresh-input rate. Agentic and long-context workloads that resend the same context on every turn can have cached tokens be the majority of all billed input, so a single blended "$X per million input tokens" estimate can be badly wrong for exactly the workloads where cost matters most. See why cached tokens usually dominate an agentic workload's bill.
Why doesn't this calculator show live Claude, GPT, or Gemini prices?
Because a wrong number is worse than no number, and API pricing changes often. This calculator does the arithmetic — multiplying your token volumes by the rates you paste in — from each provider's own current pricing page. It never fetches or asserts a vendor's price on its own.
What is a reasonable cached-input rate to use?
There is no single correct number and it varies by provider and by cache mechanism (prompt caching, context caching, implicit vs. explicit caching). Check the provider's own current pricing page for its cached/cache-read input rate, and use that — the example rates preloaded on this page are illustrative starting points, not a claim about any provider's current price.
Does this calculator upload my token counts or rates anywhere?
No. Every field runs in your browser; nothing is sent to a server. An optional anonymous local counter records which buttons get used, never the numbers you typed, and can be turned off.