Type your tokens per request, request volume and per-1M-token prices to get cost per request, day, month and year. Nothing is uploaded.
To estimate what an LLM API will cost, multiply the tokens in each request by the price per million tokens, add input and output, then multiply by requests per day and days per month. The formula is cost per request = (input tokens x input price + output tokens x output price) / 1,000,000. Worked example: 2,000 input tokens and 500 output tokens per request at 3 per 1M input and 15 per 1M output cost 0.006 + 0.0075 = 0.0135 per request. At 1,000 requests a day that is 13.50 a day, 405.00 over 30 days and 4,860.00 over 12 months. The page shows the same steps with your numbers in the box named How this was worked out.
Vendors change prices and add models often, so a table copied into this page would go stale and give you a wrong total. You type the two rates instead. Each vendor publishes them on its own pricing page: OpenAI at developers.openai.com/api/docs/pricing, Anthropic at platform.claude.com/docs/en/about-claude/pricing and Google at ai.google.dev/gemini-api/docs/pricing, as of October 2026. Copy the input and output rates for the model you use. The 3 and 15 in the boxes when the page opens are example numbers, not any vendor's price. Amounts use your browser's number format and carry no currency symbol, so use whichever currency your provider bills in.
If part of every prompt is read from a prompt cache, type that share as a percentage and the price your provider charges for cached input. The calculator splits the input tokens, prices the fresh part at the normal rate and the cached part at the cached rate. Example: 80% of 2,000 input tokens cached at 0.3 per 1M gives (400 x 3 + 1,600 x 0.3) / 1,000,000 = 0.00168 for input and 0.00918 per request, or 275.40 a month at 1,000 requests a day. A batch discount of 50% on the plain example takes 0.0135 down to 0.00675 and 405.00 a month down to 202.50. It is applied to the whole request, input and output, which is how simple batch discounts are usually described; check your provider's terms.
The result is an estimate from the numbers you type. It does not count the one-time cost of writing a cache, which some providers price separately from normal input. It does not include extra charges such as tool calls, web search, images, audio, fine-tuning or higher prices for very long prompts. It assumes every request is the same size, while real traffic varies, so use average tokens or run the page twice for a typical and a heavy request. Months are your days per month times 12 for the year line. To count tokens in a real prompt, use the Token counter page on this site; counts differ between model families.
Divide the token count by 1,000,000 and multiply by the price per million, once for input and once for output, then add them. With 2,000 input tokens at 3 per 1M and 500 output tokens at 15 per 1M the call costs 0.006 + 0.0075 = 0.0135.
On each vendor's own pricing page: developers.openai.com/api/docs/pricing for OpenAI, platform.claude.com/docs/en/about-claude/pricing for Anthropic and ai.google.dev/gemini-api/docs/pricing for Gemini (as of October 2026). Look for the input and output rates of your model.
Type the share of input that is cached, for example 80, and the cached input price per 1M tokens. With 2,000 input tokens that prices 1,600 tokens at the cached rate and 400 at the normal rate. Leave both empty for no caching.
The monthly and yearly cost if you send 10 or 100 times as many requests per day, with everything else unchanged. At 1,000 requests a day and 405.00 a month, 10x is 4,050.00 and 100x is 40,500.00. Use it to see what a launch or a busy month could cost.
The estimate assumes identical requests and only the rates you typed. Real bills vary with request size, cache writes, tool or search charges, long-prompt pricing and rounding by the provider. Treat it as a planning number and compare it with your first week of usage.
No. The numbers are processed in your browser and not uploaded. Nothing you type is saved on this device either.