Free tool · runs in your browser
AI API cost calculator
Estimate what an AI feature or chatbot will cost each month from how many requests it handles and how long they are.
Claude prices are Anthropic's list prices as of October 2026, before prompt caching or batch discounts. Check the provider's pricing page before budgeting.
Quick answer
LLM APIs charge per million tokens, with separate prices for input (what you send) and output (what the model writes). Monthly cost = requests per month × (input tokens × input price + output tokens × output price) ÷ 1,000,000. This calculator has Anthropic's Claude list prices filled in and accepts custom prices for any other model.
How to use it
- Choose a model or enter prices. Claude prices are filled in; pick Custom to enter any model's input and output price per million tokens.
- Enter your usage. Requests per day, and average input and output tokens per request.
- Read the cost. Per request, per day, per month and per year.
Frequently asked questions
How is LLM API cost calculated?
Input tokens × input price plus output tokens × output price, with prices quoted per million tokens, summed over all requests.
Why is output more expensive than input?
Generating tokens takes more computation than reading them, so providers price output several times higher.
How can I lower my AI API costs?
Use prompt caching for long repeated prompts, a smaller model where quality allows, shorter prompts and answers, and batch processing for work that is not urgent.
Are the prices current?
The Claude prices are Anthropic's published list prices as of October 2026, before discounts such as caching or batching. Check the provider's pricing page before budgeting.
Does thinking count as output tokens?
Yes. On models that think before answering, the thinking tokens are billed as output, so leave room for them in your estimate.
How many tokens does a chatbot message use?
It depends on the system prompt, the conversation history and the reply. Because the whole conversation is sent again with every message, a support chat with a long system prompt can use thousands of input tokens per message.
What is prompt caching and how much does it save?
It reuses a long prompt prefix between requests. Cached input tokens are billed at a fraction of the normal input price, which cuts the cost of bots that resend the same instructions and documents.
What is batch processing?
Sending many requests to be processed asynchronously at a discount, for work that does not need an immediate answer, such as summarising thousands of documents.
Which Claude model should I start with?
The one that meets your quality bar on real examples. Measure the cost per completed task: a cheaper model that needs more retries is not cheaper.
Does this calculator include WhatsApp or hosting costs?
No. It covers model usage only; WhatsApp conversations, servers and databases are billed separately.