LLM Cost Calculator

Paste your monthly token volumes and compare what each frontier model would cost you. The cheapest model is highlighted -- and the bottom panel shows how much you could save by routing the same workload via Inference Labs.

Defaults are typical for an established RAG workload. Cache hit rate models exact-match dedup; routing savings models switching to a cheaper tier when quality requirements allow.

Monthly cost by model

The savings number above is what Inference Labs's policy router targets. Start a free trial → -- the dashboard shows your actual cache rate and routing savings on real traffic within a week.

Need a faster diagnosis? Run the LLM Cost Leak Scanner to break out retry leakage, malformed-output reruns, and model-overkill spend.

How the math works

Cost = (input_tokens − cached) × input_price + output_tokens × output_price. Cached input is billed at $0 (we assume your cache lookup is local; on Inference Labs the cache stores embeddings + completions so a hit returns instantly with zero provider call). Routing savings apply on top of the cache: the cheaper of (cheapest model that meets quality bar) vs (your hard-coded default).