nemotron-3-super:latest

nemotron_h_moe · 123.6B · Q4_K_M

THINKING MODEL

NVIDIA NVIDIA_DGX_Spark (Cortex-X925)

122 GB · Ubuntu 24.04.5 LTS

Tested on October 3, 2026 · Submitted by Tram

Top 56% Compare

Global Score

74 /100

Good

Hardware Fit

50/100

Quality

84/100

Get this model

🦙

Ollama

ollama pull nemotron-3-super:latest

View on Ollama Library

ollama.com/library/nemotron-3-super

Get it in LM Studio

Search and download models directly from the app

🤗

Find on HuggingFace

GGUF versions & quantizations

Hardware

Machine: NVIDIA NVIDIA_DGX_Spark
CPU: Cortex-X925
Cores: 20 total (20 perf)
Frequency: 3.35 GHz
RAM: 122 GB
GPU: Device 2e12
OS: Ubuntu 24.04.5 LTS
Arch: arm64
Power Mode: performance

Performance

Tokens/sec Tokens generated per second — higher is better: 21.8
Standard deviation Variation of speed across benchmark runs — lower means more consistent: ±0.1
First chunk latency Delay before the first streamed chunk arrives from runtime: 420 ms
Time to first token How fast the model starts responding — lower is better: 2.2 s
Load time Time to load the model into memory before inference: 12.4 s
Memory usage RAM consumed during inference — percentage of total system memory: 79.5 GB (65%)
Total tokens Total number of tokens generated across all benchmark prompts: 1179
Thinking tokens (est.) Estimated internal reasoning tokens — models with "thinking" generate more tokens: ~131

Score breakdown

Speed

23/50

Time to first token

10/20

Memory

17/30

Quality

Reasoning

17/20

Coding

18/20

Instruction following

11/20

Structured output

14/15

Math

14/15

Multilingual

10/10

Category levels

Reasoning: Strong Coding: Strong Instruction Following: Adequate Structured Output: Strong Math: Strong Multilingual: Strong

Metadata

Spec version: 0.2.1
Runtime: Ollama 0.35.0
Model format: GGUF
Hardware profile: HIGH-END
Result hash: cffa0b6f9bae21b7045c1f523eff0a2c83b6df9a1e62e5505d68ce354073862b

Interpretation

Hardware fit: 50/100. Overall suitability: GOOD (Global 74/100). Category profile: Reasoning: Strong, Coding: Strong, Instruction Following: Adequate, Structured Output: Strong, Math: Strong, Multilingual: Strong.

Bench Environment

Thermal: nominal CPU load: avg 16% (peak 16%)

Run yours now

$ npm install -g metrillm@latest

$ metrillm

Requires Node 20+ and Ollama or LM Studio running

Or run without installing: npx metrillm@latest