AI Cost Calculator
Enter input and output tokens per request, pick a model, and set a daily request volume. The calculator tells you the per-request cost, the monthly spend, and how your pick compares against every other major frontier model.
Pricing last updated 2026-04-17. Compares 7 models across Anthropic, OpenAI, and Google. Client-side only; no data leaves your browser.
- Google · Gemini 2.5 Flash cheapest$6.75/mo$0.0002/req$0.075/M in · $0.3/M out
- OpenAI · GPT-5 mini$13.50/mo$0.0004/req$0.15/M in · $0.6/M out
- Anthropic · Claude Haiku 4.5$105.00/mo$0.0035/req$1/M in · $5/M out
- Google · Gemini 2.5 Pro$187.50/mo$0.0063/req$1.25/M in · $10/M out
- OpenAI · GPT-5$225.00/mo$0.0075/req$2.5/M in · $10/M out
- Anthropic · Claude Sonnet 4.6$315.00/mo$0.0105/req$3/M in · $15/M out
- Anthropic · Claude Opus 4.6$1575.00/mo$0.0525/req$15/M in · $75/M out
How to estimate tokens
A rough rule for English text is 1 token ≈ 4 characters ≈ 0.75 words. For structured outputs like JSON, expect 20 to 40% overhead for punctuation and whitespace. If you are uncertain, tokenize a representative sample of your actual prompts and responses and use the average. The wrong token estimate is the single biggest reason teams are surprised by their AI bill.
If you are building an agent loop, multiply the per-request cost by the number of LLM calls per task — agents rarely take fewer than three or four. Prompt caching, when available, typically discounts the repeated part by 90%, which is a major lever once the input is stable.