How AI model pricing works
Most large language model APIs charge per token, a unit of text that is roughly three quarters of an English word, and usually less for other European languages. Prices are quoted per million tokens and differ for input (what you send: instructions, documents, conversation history) and output (what the model writes). Output tokens are usually several times more expensive than input tokens.
Many providers also offer discounts for cached repeated input and for batch processing that does not need an immediate answer. Those can change the cost of a feature considerably.
Using the calculator
The preset rows are example prices only. Providers change prices frequently and differ by region and contract, so check current provider pricing and replace the figures before relying on the result.
- Enter the average input and output tokens per request for your use case.
- Enter requests per day and working days per month.
- Edit the per-million-token input and output prices for each model row.
- Compare the monthly and annual cost across rows.
Estimating tokens realistically
Input usually dominates in business use. A summary of a 10-page document might send 6,000 tokens and receive 500. A chatbot resends the conversation history with each message, so long chats grow quickly. Retrieval systems add the retrieved passages to every request. Measure a few real examples in a provider's console rather than guessing.
Model cost is often a small part of the total. Development, testing, monitoring, data protection review and the staff time to check outputs typically cost more than the tokens for a small firm's volumes. For confidential client data, also check where the provider processes data and whether it is excluded from training under your plan.
A worked example
An email triage assistant reads 200 messages a day of about 800 input tokens each, including instructions, and writes a 150-token summary. Over 22 working days that is about 3.5 million input and 0.66 million output tokens a month. Multiply each by the provider's current price per million tokens to get the monthly bill. Enter your own figures and prices in the calculator rather than relying on this example. Remember to add a margin for growth and retries.