Estimate your monthly LLM API bill
Loading today's prices…
| Model | Lab | Input / 1M | Output / 1M | Per month |
|---|
A request is one call to the API: a chat turn, one document summarised, one agent step. Input tokens are everything you send (the prompt, any system instructions, pasted documents). Output tokens are what the model writes back. If you only know words, multiply by about 1.3 to get tokens, or paste a real prompt into the token counter to count it exactly.
OpenAI API pricing calculator
Pick OpenAI to compare its newest paid models against each other at your traffic. OpenAI's output tokens usually cost several times its input tokens, so the output setting moves the total more than you might expect.
Claude API pricing calculator
Pick Anthropic to price Claude models side by side. Anthropic's cached-input price is a fraction of its standard input price, which matters if you send the same long instructions on every call. The estimate here uses standard prices, so it is the ceiling.
Gemini API pricing calculator
Pick Google to compare Gemini models. Google also gives a free tier for several Gemini models with its own rate limits, which this calculator does not count. The free LLM API tiers tracker lists them.
How the estimate works
Monthly cost is requests a month times (input tokens per request times the input price, plus output tokens per request times the output price), divided by one million, because labs quote prices per million tokens. Nothing else is added.
The prices are read from the LLM API price history tracker each time the page loads. That tracker takes each lab's own list price where the open-source LiteLLM price file has it, and OpenRouter's listed price otherwise, for the newest paid text models of 11 labs. If you want to know how prices have moved, the tracker shows the cheapest price for three capability levels month by month since October 2023.
Real bills often come in lower than this estimate. Prompt caching, batch APIs and volume deals all cut the price, and none of them are counted here.
FAQ
1. How does the LLM API cost calculator work out the monthly cost?
Monthly cost = requests a month x (input tokens per request x input price + output tokens per request x output price) / 1,000,000. Prices are each model's standard list price per million tokens.
2. Where do the prices come from?
From Gravity's LLM API price history tracker. Each model uses the lab's own list price where the open-source LiteLLM price file matches it, and OpenRouter's listed price otherwise. The price source is shown for every row in the downloadable CSV.
3. Does the estimate include cached input, batch or volume discounts?
No. The total uses standard list prices for input and output tokens. Prompt caching, batch APIs and negotiated discounts can lower a real bill a lot, so treat the result as the undiscounted ceiling.
4. How many tokens is a word?
For English text, OpenAI's rule of thumb is that one token is about four characters, or about three quarters of a word, so 1,000 words is roughly 1,300 tokens. Other languages and code usually take more tokens per word.
5. Which models are in the calculator?
The newest paid text models from 11 labs: OpenAI, Anthropic, Google, xAI, DeepSeek, Mistral, Alibaba (Qwen), Moonshot AI, Z.ai, MiniMax and Cohere, up to five per lab, as listed in the price tracker on the day you open the page.
6. Can I reuse the price data?
Yes. The prices behind the calculator are CC BY 4.0. Download current-prices.csv and credit Gravity with a link to the LLM API price history page.