AI Desk · Model Identity · Evidence object
Tokenizer signature tk_59ca250f
An exact tokenizer signature: every model listed below returned the same marginal prompt-token count on all 50 test strings. Sharing a signature is strong evidence of a shared vocabulary. This signature sits in the Gemini · group A group.
- Members
- 20
- First observed
- 21 August 2026
- Last observed
- 3 September 2026
- Group
- Gemini · group A
Models with this signature
Google: Nano Banana (Gemini 2.5 Flash Image)google/gemini-2.5-flash-imageGoogleGoogle: Gemini 2.5 Flash Litegoogle/gemini-2.5-flash-liteGoogleGoogle: Gemini 2.5 Progoogle/gemini-2.5-proGoogleGoogle: Gemini 2.5 Pro Preview 06-05google/gemini-2.5-pro-previewGoogleGoogle: Gemini 2.5 Pro Preview 05-06google/gemini-2.5-pro-preview-05-06GoogleGoogle: Gemini 3 Flash Previewgoogle/gemini-3-flash-previewGoogleGoogle: Nano Banana 2 (Gemini 3.1 Flash Image Preview)google/gemini-3.1-flash-image-previewGoogleGoogle: Gemini 3.1 Flash Litegoogle/gemini-3.1-flash-liteGoogleGoogle: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)google/gemini-3.1-flash-lite-imageGoogleGoogle: Gemini 3.1 Flash Lite Previewgoogle/gemini-3.1-flash-lite-previewGoogleGoogle: Gemini 3.1 Pro Previewgoogle/gemini-3.1-pro-previewGoogleGoogle: Gemini 3.1 Pro Preview Custom Toolsgoogle/gemini-3.1-pro-preview-customtoolsGoogleGoogle: Gemini 3.5 Flashgoogle/gemini-3.5-flashGoogleGoogle: Gemini 3.5 Flash Litegoogle/gemini-3.5-flash-liteGoogleGoogle: Gemini 3.6 Flashgoogle/gemini-3.6-flashGoogleGoogle: Gemini 3.7 Flashgoogle/gemini-3.7-flashGoogleGoogle: Gemini 3.8 Flashgoogle/gemini-3.8-flashGoogleGoogle: Gemma 3 12Bgoogle/gemma-3-12b-itGoogleGoogle: Gemma 3 27Bgoogle/gemma-3-27b-itGoogleGoogle: Gemma 3 4Bgoogle/gemma-3-4b-itGoogle
Template variants of the same tokenizer
These signatures differ from this one by a uniform one-token shift on the affected strings: the same vocabulary behind a different chat template.
The measured vector
Marginal prompt-token cost per string, as measured on Google: Nano Banana (Gemini 2.5 Flash Image)All 50 rows
English and whitespace
en-prose14
en-long16
spaces-203
spaces-604
tabs-203
newlines-203
mixed-ws9
Digits
digits-99
digits-1212
digits-3030
digits-sep12
float-long22
CJK
zh-common10
zh-long12
zh-rare25
ja-kana7
ja-kanji9
ko11
Other scripts
ru12
ar10
he13
hi8
th10
el13
Emoji
emoji-basic5
emoji-skin10
emoji-zwj-family7
emoji-zwj-x321
emoji-flags8
emoji-prof11
Rare Unicode
math26
boxdraw17
combining10
cjk-ext-b16
surrogates27
zalgo22
rtl-mix5
Code
py-code28
py-indent18
json28
html14
regex44
camel6
snake11
Repetition and encodings
rare-word-x525
repeat-tok21
base6434
hex16
url23
uuid36
Method: tokenizer fingerprinting, v1-50.