275 models on 9 platforms, prices as of Oct 19, 2026.
Pick what you are building, a chatbot, a coding assistant, document summaries or customer support, and the calculator fills in a typical day for it: how many requests, how many words go in and how many come back. Change any number and the card shows what it costs a month on the cheapest model, or on any model you pick. Below it every Claude, GPT, Gemini, Grok, Mistral and DeepSeek model is listed cheapest first, with its cost a month, a day and a request. Prices are read from the public LiteLLM model table every day. Computed in your browser, nothing sent anywhere.
How it works
Pick what you are building and see what it costs a month on every AI model
AI companies charge for every request your app sends. The bill comes down to three things: how many requests a day, how much text goes in (your message, the instructions and the earlier conversation) and how much comes back. Pick what you are building and the calculator fills in a typical day for it; change any number to match your own. The card opens on the cheapest model; pick any other model there, or click a row in the table, to see its bill.
Vendors count text in tokens, pieces of words, about 750 words to a thousand tokens in English, so the calculator asks in words and converts them for you. More Options holds the rest: the share of each request that repeats, such as fixed instructions, which most models bill much cheaper; the days your app runs in a month; and a monthly budget, which adds a column with how many requests a day it pays for on each model. Prices are read from the public LiteLLM model table every day, and everything is computed in your browser.
Questions
- Which model should I pick?
- The cheapest row is not always the right one: small models handle simple questions well but struggle with long or hard tasks. A common path is to start with a mid range model, then try a cheaper one on your real requests and keep it if the answers hold up.
- How many words is a request?
- Count everything the model reads: your message, the fixed instructions and, in a chat, the earlier messages it has to remember. A chat message is often 500 to 3,000 words in and 100 to 500 words back. Code and languages other than English take more tokens per word.
- Does this include subscriptions like ChatGPT Plus?
- No. It prices the API, which is what an app pays per request. A chat subscription is a flat monthly fee for one person; enter that fee as your budget under More Options to see how many requests a day it would buy at API prices.
- Are the prices current?
- They are read once a day from the public model price table of LiteLLM, an open source project the AI usage tools read too, so a new model or a price change shows up within a day. The result is an estimate; check the vendor pricing page before a large commitment.
