Calculator
Prompt Cost Estimator.
Pick a model, set token volume, get monthly and annual cost. Last updated 2026-05.
Inputs
Cost per call
$0.0120
Cost per day
$12.00
Monthly
$360
~30 days × calls/day
Annual
$4,320
Frequently asked questions
How is monthly cost calculated?
Cost = (input tokens × input price/M) + (output tokens × output price/M), then × calls per day × 30 days. Token counts are approximate — actual cost depends on tokenization specifics.
How big is an average prompt?
Heuristic: 1 token ≈ 0.75 words in English. A 500-word system prompt + 100-word user message ≈ 800 input tokens. Output varies hugely — short JSON ≈ 50 tokens, long-form ≈ 500–2,000.
Why is self-hosted Llama "estimated"?
Self-hosting has no per-token pricing — your cost is GPU time. We approximate $0.30/M tokens both directions based on a typical H100 deployment at common throughput. Real cost depends on hardware utilization.
Are these prices current?
Last updated 2026-05. Provider pricing changes frequently — always confirm with the official pricing page before committing budget.