LLM API cost calculator

The projection averages your token shape over a month of traffic, so it cannot see bursts, retries, failed requests, or tier switches at a context boundary. Regional surcharges, priority modes, and image or audio pricing are out of scope. Cache economics are only as good as your hit-rate guess: on Anthropic models a 5-minute cache write costs 1.25 times the input price, so a low hit rate can make caching cost more than sending everything at list price, and the savings figure below goes negative to say so.