AI Desk · Model Identity · Compare

Google: Gemini 3.7 Flash · NVIDIA: Nemotron 3.5 Content Safety (free)

Identity similarity: lowDivergent fingerprints · 22/50These models tokenise differently. Divergence on the high-information strings (CJK, emoji, digits) separates unrelated vocabularies within a handful of samples.

Google: Gemini 3.7 Flash recordNVIDIA: Nemotron 3.5 Content Safety (free) recordOpen in the interactive bench

Tokenizer fingerprint

22 of 50 mutually clean strings agree
Test stringGoogle: Gemini 3.7 FlashNVIDIA: Nemotron 3.5 Content Safety (free)Verdict
en-prose1414match
en-long1615differs
spaces-2032differs
spaces-6042differs
tabs-2032differs
newlines-2032differs
mixed-ws92differs
digits-9910differs
digits-121213differs
digits-303031differs
digits-sep1213differs
float-long2223differs
zh-common1011differs
zh-long1214differs
zh-rare2526differs
ja-kana77match
ja-kanji910differs
ko1111match
ru1212match
ar109differs
he1312differs
hi88match
th1010match
el1313match
emoji-basic55match
emoji-skin1010match
emoji-zwj-family78differs
emoji-zwj-x32122differs
emoji-flags89differs
emoji-prof1112differs
math2626match
boxdraw1718differs
combining1010match
cjk-ext-b1617differs
surrogates2728differs
zalgo2222match
rtl-mix55match
py-code2826differs
py-indent1812differs
json2828match
html1414match
regex4444match
camel66match
snake1111match
rare-word-x52525match
repeat-tok2120differs
base643434match
hex1616match
url2323match
uuid3637differs

Values are the marginal prompt-token cost of each string with the model's fixed overhead subtracted. An x marks a row excluded as corrupt for that model.

API surface

The declared serving contracts differ
FieldGoogle: Gemini 3.7 FlashNVIDIA: Nemotron 3.5 Content Safety (free)Verdict
Context length1048576128000× differs
Max output tokens655368192× differs
Supported parametersinclude_reasoning, max_tokens, reasoning, reasoning_effort, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_pinclude_reasoning, max_tokens, reasoning, seed, temperature, top_p× differs
Defaults{}{"temperature":null,"top_p":null,"top_k":null,"frequency_penalty":null,"presence_penalty":null,"repetition_penalty":null}× differs
Reasoning contract{"mandatory":true,"default_enabled":true,"supported_efforts":["high","medium","low"],"default_effort":"medium"}{"mandatory":false,"default_enabled":true}× differs
Modalitiestext+image+file+audio+video->texttext+image->text× differs

Method: tokenizer fingerprinting, v1, 50 strings. Raw data on the data page. This page is static and dated; the interactive bench compares any two of the 422 catalogue models.