Concurrency-Aware LLM Cost Explorer

Static per-token calculators hide load assumptions. This browser-only calculator makes those assumptions explicit using Little's Law and throughput-aware costing.

Read the paper Open-source implementation

Inputs

Derived estimates

Approximate in-flight concurrency (Little's Law)
-
Effective cost / million output tokens
-
Fully utilized cost / million output tokens
-
Underutilization multiplier
-
Estimated utilization
-

Warnings

Provenance: editable fields are user-provided values. Presets are illustrative references, not live measurements. Derived values are labelled as estimates.

Illustrative cost versus offered load

Illustrative cost versus offered load A line chart of cost per million output tokens as offered request rate increases from low to high load based on the entered assumptions.
This chart is a simulation from current inputs and does not measure a running deployment.