AI Desk · Model Identity

Tokenizer Atlas

Explore the tokenizer-family structure underlying the AI model ecosystem: which models read text with the same vocabulary, where the families sit, and which entries share ancestry their labels never mention.

Models measured: 158
Distinct fingerprints: 46
Tokenizer groups: 35
Groups with 2+ models: 21
As at: 22 August 2026

How to read this page: a tokenizer is the fixed vocabulary a model chops text with, and it cannot change without retraining, so it works like a fingerprint. Models in one group below chop all fifty of our test strings identically, which means they share a vocabulary, and usually an ancestry.

The constellation

One star cluster per tokenizer group; size follows membership

Hover a cluster for its family; click to jump to its members. The full listing below carries everything the canvas shows, so nothing depends on it.

Family explorer

One card per shared vocabulary; every model listed on a card chops text identically

Qwenfg_5410b38e · 27 models · 1 signature

and 19 more

Mistralfg_e0b4c23e · 15 models · 3 signatures

and 7 more

DeepSeekfg_869dc1cf · 12 models · 2 signatures

and 4 more

Llama3fg_066351de · 11 models · 3 signatures

and 3 more

Qwenfg_c5c0a8ed · 11 models · 2 signatures

and 3 more

GPTfg_d1411c9d · 10 models · 1 signature

and 2 more

Geminifg_a15512c7 · 8 models · 2 signatures

MiniMaxfg_9f84b5fb · 7 models · 1 signature

Z.aifg_1d05586d · 4 models · 2 signatures

InclusionAIfg_fc5a4585 · 4 models · 1 signature

Tencentfg_2b541395 · 4 models · 1 signature

ByteDance Seedfg_25bf52d0 · 4 models · 1 signature

OpenAI cl100kfg_ba918336 · 3 models · 1 signature

Llama4fg_20b9d005 · 3 models · 2 signatures

Qwenfg_e60ca14a · 3 models · 2 signatures

Llama2fg_5348f9f6 · 2 models · 2 signatures

Upstagefg_865d84ce · 2 models · 1 signature

Novafg_76e424dd · 2 models · 1 signature

Mistralfg_ddc193e9 · 2 models · 1 signature

Gemmafg_ba03436b · 2 models · 1 signature

Z.aifg_3eaaf3e0 · 2 models · 1 signature

One of a kind, so far

A unique EXACT fingerprint does not take a model out of its family: the nearest relative, and how close it comes, is shown on each row

A model lands here when its exact 50-number fingerprint has no twin yet, which usually means a vocabulary revision inside its own family (Qwen3.8 sits close to the Qwen line without matching it exactly) or relatives that are still unmeasured.

Reading the Atlas

A group gathers every model whose measured fingerprints are the same vocabulary: exact signature matches, plus signatures one uniform template shift apart. Group membership is lineage evidence, not release identification, and the methodology records exactly how much the fingerprints can distinguish. The Atlas's working face is the unknowns queue: what is still unplaced, and what measurement has already placed.