YFarmX

AI Desk · Model Identity · Compare

NVIDIA: Nemotron 3 Nano Omni (free) · TheDrummer: UnslopNemo 12B

Identity similarity: lowDivergent fingerprints · 31/50These models tokenise differently. Divergence on the high-information strings (CJK, emoji, digits) separates unrelated vocabularies within a handful of samples.

NVIDIA: Nemotron 3 Nano Omni (free) recordTheDrummer: UnslopNemo 12B recordOpen in the interactive bench

Tokenizer fingerprint

31 of 50 mutually clean strings agree
Test stringNVIDIA: Nemotron 3 Nano Omni (free)TheDrummer: UnslopNemo 12BVerdict
en-prose1414match
en-long1818match
spaces-2033match
spaces-6033match
tabs-2044match
newlines-2077match
mixed-ws77match
digits-9910differs
digits-121213differs
digits-303031differs
digits-sep1213differs
float-long2223differs
zh-common1617differs
zh-long2020match
zh-rare2021differs
ja-kana1112differs
ja-kanji1313match
ko1312differs
ru1212match
ar98differs
he1010match
hi1414match
th1717match
el1313match
emoji-basic2019differs
emoji-skin4039differs
emoji-zwj-family1918differs
emoji-zwj-x35756differs
emoji-flags3231differs
emoji-prof3534differs
math3333match
boxdraw3030match
combining1414match
cjk-ext-b1616match
surrogates5353match
zalgo3636match
rtl-mix66match
py-code2525match
py-indent1515match
json2525match
html1617differs
regex4950differs
camel66match
snake77match
rare-word-x53535match
repeat-tok4040match
base643939match
hex1919match
url1919match
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
FieldNVIDIA: Nemotron 3 Nano Omni (free)TheDrummer: UnslopNemo 12BVerdict
Context length2560001024000× differs
Max output tokens65536819200× differs
Supported parametersinclude_reasoning, max_tokens, reasoning, seed, temperature, tool_choice, tools, top_pfrequency_penalty, logit_bias, logprobs, max_tokens, presence_penalty, repetition_penalty, seed, stop, structured_outputs, temperature, top_k, top_logprobs, top_p× differs
Defaults{"temperature":0.6,"top_p":0.95,"top_k":null,"frequency_penalty":null,"presence_penalty":null,"repetition_penalty":null}{}× differs
Reasoning contract{"mandatory":false,"default_enabled":true,"supports_max_tokens":true}× differs
Modalitiestext+image+audio+video->texttext->text× differs

Method: tokenizer fingerprinting, v1, 50 strings. This page is static and dated; the interactive bench compares any two of the 486 catalogue models.