Skip to content
SFAI
SFAI / PRACTICAL GUIDE

How to read AI model usage rankings

Observed usage measures a source’s traffic, not the entire global AI market.

Define the measurement

OpenRouter token observations and modality request observations are different measures. A token total cannot be added to a request total. A provider marketplace is also only one traffic population; popularity on that marketplace does not establish worldwide usage or model quality.

Treat missing rows correctly

Published ranking feeds can expose only named leaders and an aggregated Others bucket. A missing model row is unknown, not zero. Our model charts leave gaps, and daily gainers require comparable observations for the same variant. Free variants and other configurations retain distinct source identities.

Avoid double counting

An aggregate marketplace feed can overlap with application or provider reports. Adding those numbers without evidence that their populations are disjoint can count the same requests twice. NATDAQ keeps sources and units separate and attributes observations to their original reporting window.

Apply the method to current data

Browse current source-qualified model profiles · Review sources and publication limits