NVIDIA: Nemotron 3.5 Content Safety (free)

nvidia/nemotron-3.5-content-safety:free · NVIDIANot yet established

Tokenizer group (measured): NVIDIA
Entered catalogue: 4 June 2026
Last tested: 21 August 2026
Evidence confidence: Low

Why low: inherited from nvidia/nemotron-3.5-content-safety: a serving variant shares its base model's tokenizer.

Evidence

✓ inherited from nvidia/nemotron-3.5-content-safety: a serving variant shares its base model's tokenizer

Compare with NVIDIA: Nemotron 3.5 Content Safety

Report a correction

This entry is a serving variant of NVIDIA: Nemotron 3.5 Content Safety. A variant shares its base model's tokenizer by construction, so the measurement lives on the base record and this page inherits it.

Identity statement

NVIDIA: Nemotron 3.5 Content Safety (free) is NVIDIA's model, and its measured tokenizer sits in NVIDIA's own signature group alongside its sibling models. No other maker's measured model shares it so far.

What we checked

Four different kinds of evidence, and what each one showed

We checked	What we found	Where it came from
How it counts tokens	Counted cleanly, and no other model in the catalogue counts the same way	tk_9a840805
How its API is set up	A combination of settings no other listing in the catalogue uses	api_f125dce8
How much it can read and write	Reads up to 128,000 tokens at once, writes up to 8,192	its own listing
Whether it thinks before answering	Reasons when asked to	its own listing
Who served our requests	Nvidia	we measured it, 21 August 2026

Closest measured models

Agreement across the mutually clean test strings

NVIDIA: Nemotron 3.5 Content Safetynvidia/nemotron-3.5-content-safety43/50 · 86%Google: Nano Banana 2 (Gemini 3.1 Flash Image)google/gemini-3.1-flash-image21/44 · 48%Google: Gemini 3.7 Flashgoogle/gemini-3.7-flash22/50 · 44%Google: Gemma 3 4Bgoogle/gemma-3-4b-it22/50 · 44%Google: Gemma 3 12Bgoogle/gemma-3-12b-it22/50 · 44%Google: Gemma 3 27Bgoogle/gemma-3-27b-it22/50 · 44%

Click a row to open the full comparison.

The measured fingerprint

What each test string cost this model, in tokens

We sent NVIDIA: Nemotron 3.5 Content Safety (free) fifty short pieces of text and recorded what each one cost it in tokens. The bars below are those costs. Two models built on the same tokenizer produce the same bars; a model built on a different one produces a different set, which is what makes this a fingerprint.

All 50 rows

English and whitespace

en-prose14

en-long15

spaces-202

spaces-602

tabs-202

newlines-202

mixed-ws2

Digits

digits-910

digits-1213

digits-3031

digits-sep13

float-long23

CJK

zh-common11

zh-long14

zh-rare26

ja-kana7

ja-kanji10

ko11

Other scripts

ru12

ar9

he12

hi8

th10

el13

Emoji

emoji-basic5

emoji-skin10

emoji-zwj-family8

emoji-zwj-x322

emoji-flags9

emoji-prof12

Rare Unicode

math26

boxdraw18

combining10

cjk-ext-b17

surrogates28

zalgo22

rtl-mix5

Code

py-code26

py-indent12

json28

html14

regex44

camel6

snake11

Repetition and encodings

rare-word-x525

repeat-tok20

base6434

hex16

url23

uuid37

Overhead subtracted: 462 prompt tokens.

Declared record

What the catalogue claims about this model

NVIDIA Nemotron 3.5 Content Safety is a compact 4B-parameter multimodal guardrail model from NVIDIA, fine-tuned from Google Gemma-3-4B. It moderates both inputs to and responses from LLMs and VLMs, accepting...

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

Supported parameters

include_reasoningmax_tokensreasoningseedtemperaturetop_p

defaults: {"temperature":null,"top_p":null,"top_k":null,"frequency_penalty":null,"presence_penalty":null,"repetition_penalty":null}

Catalogue entry Weights on Hugging Face

History

Every observation, kept as taken

4 June 2026Enters the OpenRouter catalogue with no declared family.
21 August 2026Fingerprinted in the YFarmX catalogue sweep · 50 of 50 strings measured clean.
21 August 2026YFarmX assessment: consistent with the NVIDIA family, low confidence.