How much does an LLM API cost per month?

Updated Oct 9, 2026

Providers price language models per million tokens, separately for input and output. A monthly estimate needs only three numbers.

The formula

Monthly cost = input tokens × input price + output tokens × output price, with the tokens in millions.

For example, a support bot answering 20,000 questions a month, each with about 1,500 tokens of prompt and history and a 300-token answer:

• input: 20,000 × 1,500 = 30 million tokens;
• output: 20,000 × 300 = 6 million tokens.

At $0.20 input and $0.80 output per million that is 30 × 0.20 + 6 × 0.80 = $10.80 a month. The same traffic on a model at $3 and $15 costs $180.

How to count your tokens

• One token is about three quarters of an English word; other languages often need more tokens for the same text.
• Count the whole prompt: system instructions, retrieved documents and the chat history are sent again with every request.
• Reasoning models also bill their hidden "thinking" as output tokens, often several times the visible answer.

Traps

• Cache prices. Many providers charge less for a repeated prompt prefix (cached input). If your prompts share a long system part, the real bill can be much lower.
• Minimums and subscriptions. Some offers add a monthly fee or a minimum spend.
• Free tiers. They usually have strict rate limits; check that your traffic fits before counting on them.

Let the site do it

On the Models page, type your monthly input and output (for example 30 and 6, in millions) into "Your cost/mo": every offer then shows its monthly bill, and the list sorts by it. The cheapest today:

Live data: this table is built from today's prices.