What this calculator answers

Large language model API costs scale with two numbers: input tokens (what you send to the model) and output tokens (what it sends back). Different providers and tiers price these differently, and the cost of a heavy production workload can easily run into thousands of dollars a month before anyone notices. This tool estimates your monthly spend across the major providers given your prompt volume, average prompt length, and average response length.

How the math works

A token is roughly four characters of English text, or about three-quarters of a word. Each provider publishes input and output prices per million tokens. The tool takes your daily request volume, multiplies by your average input and output token counts per request, scales to a monthly total, and multiplies by each provider's input and output rates. The result is a comparison grid of estimated monthly cost.

When to use it

• You are launching a new feature backed by an LLM and need to forecast monthly API spend before committing to a budget.
• You are comparing GPT-class, Claude-class, Gemini-class, and open-weights options for a specific workload.
• Your API bill suddenly spiked and you want to model where the cost is concentrated (heavy input prompts versus long outputs).
• You are negotiating an enterprise contract and want a defensible internal estimate of your usage in dollar terms.

Common mistakes

• Forgetting that system prompts count against input tokens on every request. A 2,000 token system prompt called a million times a month adds two billion tokens of input cost.
• Underestimating output length. Models are chatty by default. A 200 word average output is closer to 270 tokens than 200.
• Ignoring tool-use and function-calling overhead. Each tool definition and tool response counts as tokens. A heavy agentic workflow can double your token count per user request.
• Pricing only the model you are testing with. Most production systems combine a cheap model for routing with a premium model for hard cases. Always model the blended cost.

A worked example

You expect 5,000 user requests per day, each with about 1,500 input tokens (including a 600 token system prompt) and 400 output tokens. That is 7.5 million input tokens and 2 million output tokens daily, or roughly 225 million input and 60 million output per 30-day month. On a model priced at 3 dollars per million input and 15 dollars per million output, that is 675 dollars of input plus 900 dollars of output, or 1,575 dollars per month. Swap to a model at 0.25 dollars input and 1.25 dollars output and the same workload costs 131 dollars per month.

Frequently asked questions