Enter your actual usage pattern — requests per day, and roughly how many tokens go in and come out per request — to get a real monthly estimate, not just a per-token price.
Cost per request = (uncached input tokens × input price + cached input tokens × cached-input price + output tokens × output price) ÷ 1,000,000, using the model’s prices from our dataset. Per day = per request × requests per day; per month = per day × 30; per year = per day × 365.
When a model publishes no cached-input price, cached tokens are charged at the full input price instead of a guessed discount. Prices are standard pay-as-you-go rates in US dollars: batch discounts, regional surcharges and long-context surcharges are not applied (the notes in the comparison table say where they exist).
A support chatbot answers 1,000 requests a day, each with 1,500 input tokens and 400 output tokens, on Anthropic Sonnet 5.5 ($2 input and $10 output per million tokens, checked 2026-10-25).
Per request: 1,500 × $2 ÷ 1M + 400 × $10 ÷ 1M = $0.0070. That is $7.00 a day, $210 a month and $2,555 a year.
If 60% of the input is a repeated system prompt read from cache at $0.1 per million, the month drops to $159, a 24% saving. The same traffic on OpenAI GPT-6 Luna costs $10.50 a month.
Every price comes from the provider’s own pricing page:
Go deeper: What a token is, and how to estimate tokens for your text · Prompt caching and batch discounts, provider by provider · How to estimate the monthly cost of a chatbot, RAG app or agent · Why output tokens cost more than input tokens · Price history and changelog
You pay separately for input tokens (what you send) and output tokens (what the model writes), each priced per million tokens. Cost per request = input tokens × input rate + output tokens × output rate.
For English text, one token is roughly 4 characters or about 0.75 words, so 1,000 words is around 1,300 tokens. Code and non-English text usually take more tokens.
Generating text takes more computation than reading it, so output costs 2 to about 8 times as much as input across the models tracked here, most often 5 times. Shorter answers are often the biggest saving.
Some providers charge much less for input that repeats across requests, such as a long system prompt or shared document. With a lot of repeated context it can cut input cost by 70–90%.
Free to use on your own site, docs, or blog post — paste this code where you want it to appear.