LLM pricing

AI API providers charge per token — a unit roughly equal to three-quarters of a word — with separate rates for input (the prompt) and output (the generated reply). Output tokens cost more because they require sequential generation steps the model can't parallelize. Prices are quoted per million tokens and vary by orders of magnitude across models. The hard part for builders: total cost depends on context length (long prompts multiply fast), call volume, and how much of the work is input versus output — a naive estimate rarely survives contact with real traffic.