tokenstat
tokenstat

Model catalog

Usage cost calculator

Plug in how many million tokens you expect, pick a model (or enter rates), and see list-rate dollars. Assumptions are yours.

Models · AI plans · Install tokenstat to measure real sessions

Using cheapest list rates for NVIDIA: llama-3_2-nemoretriever-300m-embed-v1.

INPUT
$30
10 MTok × $3
OUTPUT
$30
2 MTok × $15
CACHE READ
$12
40 MTok × $0.3
CACHE WRITE
$7.50
2 MTok × $3.75

List-rate total for this shape

$79.50

Million tokens × published $/MTok. Cache rates use the cheapest listed cache field when the model is known. Otherwise the custom rates you entered.

Measure the tokens behind your work

The app reads the session logs your harnesses already write and lays them out by model, project, and day. Track the tokens behind your work and the list-rate value your subscription covered.

CLI for macOS, Linux, and Windows