Observed arrival · 2026-08-27
Modelbehind: The Tokenizer Fingerprint Test
An open tool that tests whether an API endpoint’s claimed model matches its likely model family by analyzing tokenizer usage.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
Field notes
The tool accepts an endpoint, its advertised model name, and a temporary API key, then issues test requests and compares token usage with a tokenizer fingerprint database. It offers a normal mode of about six requests and a stricter mode of about 26. The site explicitly limits the result to a likely model family: shared vocabularies such as o200k can prevent it from separating otherwise different models.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue