YFarmXIntelligence for an accelerating worldBriefings

AI Desk · Model Identity

NVIDIA: Nemotron Nano 12B 2 VL (free)

nvidia/nemotron-nano-12b-v2-vl:free · NVIDIANot yet established
Tokenizer lineage (measured)
Mistral
Evidence confidence
Low
Why low: partial agreement (40 of 47 strings) with nvidia/nemotron-nano-9b-v2:free only.
Declared tokenizer tag
Other
Entered catalogue
28 October 2025
Last tested
21 August 2026
Measurement
Measured
Evidence
  • ✓ partial agreement (40/47) with nvidia/nemotron-nano-9b-v2:free
Compare with NVIDIA: Nemotron Nano 9B V2 (free)

Download JSONReport a correction

Identity statement

NVIDIA: Nemotron Nano 12B 2 VL (free) is NVIDIA's model. Its measured tokenizer sits with the Mistral group, which is evidence that it builds on the Mistral tokenizer. Tokenizer reuse is normal engineering, open vocabularies travel between labs, and this says nothing against NVIDIA's authorship of the model itself.

Evidence stack

Signals of different kinds, weighed together
SignalResultReference
Tokenizer signatureMeasured with 3 corrupt rows excluded–
API surfaceNo other entry shares this exact contractapi_22e5c378
Context and output128,000 context · 128,000 max outputdeclared
Reasoning contractOptionaldeclared
Serving providers observedNvidiameasured 21 August 2026

Closest measured models

Agreement across the mutually clean test strings
NVIDIA: Nemotron Nano 9B V2 (free)nvidia/nemotron-nano-9b-v2:free40/47 · 85%NVIDIA: Nemotron 3 Nano Omni (free)nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free34/47 · 72%Mistral: Mistral Nemomistralai/mistral-nemo32/47 · 68%NVIDIA: Nemotron 3 Nano 30B A3Bnvidia/nemotron-3-nano-30b-a3b32/47 · 68%NVIDIA: Nemotron 3.5 Lightningnvidia/nemotron-3.5-lightning32/47 · 68%NVIDIA: Nemotron 3 Supernvidia/nemotron-3-super-120b-a12b32/47 · 68%

Click a row to open the full comparison.

The measured fingerprint

Marginal prompt-token cost of each test string, grouped by script
All 50 rows, 3 excluded as corrupt

English and whitespace

en-prose13
en-long17
spaces-202
spaces-602
tabs-202
newlines-202
mixed-ws2

Digits

digits-99
digits-1212
digits-3030
digits-sep12
float-long22

CJK

zh-commonx
zh-long20
zh-rare20
ja-kana11
ja-kanji13
ko13

Other scripts

ru12
ar9
he10
hi14
th17
el13

Emoji

emoji-basic20
emoji-skin40
emoji-zwj-family19
emoji-zwj-x3x
emoji-flags32
emoji-prof35

Rare Unicode

math33
boxdraw30
combining14
cjk-ext-b16
surrogates53
zalgox
rtl-mix6

Code

py-code23
py-indent12
json24
html15
regex48
camel6
snake7

Repetition and encodings

rare-word-x535
repeat-tok40
base6438
hex19
url19
uuid36

An x marks a row excluded as corrupt (caching or provider interference detected during measurement). Overhead subtracted: 15 prompt tokens.

Declared record

What the catalogue claims about this model

NVIDIA Nemotron Nano 2 VL is a 12-billion-parameter open multimodal reasoning model designed for video understanding and document intelligence. It introduces a hybrid Transformer-Mamba architecture, combining transformer-level accuracy with Mamba’s...

Modalities
text+image+video->text
Declared tokenizer
Other
Prompt price
$0 / M tokens
Completion price
$0 / M tokens

Supported parameters

include_reasoningmax_tokensreasoningseedtemperaturetool_choicetoolstop_p

defaults: {"temperature":null,"top_p":null,"frequency_penalty":null}

Catalogue entryWeights on Hugging Face

History

Every observation, kept as taken
  • 28 October 2025Enters the OpenRouter catalogue with no declared family.
  • 21 August 2026Fingerprinted in the YFarmX catalogue sweep · 47 of 50 strings measured clean.
  • 21 August 2026YFarmX assessment: consistent with the Mistral family, low confidence.

Cite this page as the evidence record for NVIDIA: Nemotron Nano 12B 2 VL (free): the URL is stable, measurements are dated, and revisions append to the history above. Method: tokenizer fingerprinting, v1, 50 strings. The raw data behind every figure is on the data page.