Most AI app ideas look cheap in a demo and become expensive at scale. A prototype with 100 users may cost almost nothing. The same product with 10,000 active users, long chat history, large prompts and expensive models can become a serious monthly bill. This calculator estimates the real LLM cost before you build β€” users, calls, tokens, model choice and simple optimisations like caching or batching.

β€”
per month
β€”per user / month
β€”per year
β€”LLM calls / month

Cheapest models for this workload

Same usage, every model ranked by monthly cost.

ModelCost / monthPer user
⚠️ Estimate using reference prices (June 2026) and list rates. Real bills vary with caching, batching, region, tiers and add-ons. Type your real token counts for the closest figure.

Build your AI app fast with v0 β†’
Referral link β€” v0 by Vercel turns a prompt into a working app UI. Then estimate its running cost above.

How to read this

Your AI bill is mostly users Γ— calls Γ— tokens Γ— price. The biggest levers, in order: which model (small vs frontier is often 20–50Γ— cheaper), output length (output costs 3–5Γ— input β€” cap it), and input length (trim system prompts and retrieved context). Caching repeated context and batching non-urgent jobs cut more.

Picking a model? Compare them head-to-head on the AI API cost calculator, or read GPT vs Claude vs Gemini. Building RAG or a chatbot? Those have their own cost drivers (embeddings, vector DB, retrieval size) β€” guides coming.