qwen3:4b

qwen3 · 4.0B · Q4_K_M

NVIDIA NVIDIA Jetson Orin Nano Engineering Reference Developer Kit Super (Cortex-A78AE)

7 GB · Ubuntu 22.04.5 LTS

Tested on June 26, 2026

Top 96% Compare

Global Score

32 /100

Not Rec.

Hardware Fit

98/100

Quality

4/100

Get this model

🦙

Ollama

ollama pull qwen3:4b

View on Ollama Library

ollama.com/library/qwen3

Get it in LM Studio

Search and download models directly from the app

🤗

Find on HuggingFace

GGUF versions & quantizations

Hardware

Machine: NVIDIA NVIDIA Jetson Orin Nano Engineering Reference Developer Kit Super
CPU: Cortex-A78AE
Cores: 6 total (6 perf)
Frequency: 1.17 GHz
RAM: 7 GB
GPU: Integrated / Unknown
OS: Ubuntu 22.04.5 LTS
Arch: arm64
Power Mode: balanced

Performance

Tokens/sec Tokens generated per second — higher is better: 17.9
Standard deviation Variation of speed across benchmark runs — lower means more consistent: ±0.0
First chunk latency Delay before the first streamed chunk arrives from runtime: 950 ms
Time to first token How fast the model starts responding — lower is better: 950 ms
Load time Time to load the model into memory before inference: 9.7 s
Memory usage RAM consumed during inference — percentage of total system memory: 3.0 GB (40%)
Total tokens Total number of tokens generated across all benchmark prompts: 1320

Score breakdown

Speed

50/50

Time to first token

20/20

Memory

28/30

Quality

Reasoning

4/20

Coding

0/20

Instruction following

0/20

Structured output

0/15

Math

0/15

Multilingual

0/10

Category levels

Reasoning: Poor Coding: Poor Instruction Following: Poor Structured Output: Poor Math: Poor Multilingual: Poor

Metadata

Spec version: 0.2.1
Runtime: Ollama 0.30.10
Model format: GGUF
Hardware profile: ENTRY
Result hash: 7a2fee519e667f0db7823f1aff5eb9d64ae61f2e9058e06077e92343f4df6a08

Interpretation

Hardware fit: 98/100. Overall suitability: NOT RECOMMENDED (Global 32/100). Category profile: Reasoning: Poor, Coding: Poor, Instruction Following: Poor, Structured Output: Poor, Math: Poor, Multilingual: Poor. Warning: model produced very low accuracy on quality tasks — results may be unusable despite good hardware performance.

Warnings

Model produced very low accuracy on quality tasks — results may be unusable despite good hardware performance.

Bench Environment

Thermal: nominal Swap delta: +0.1 GB CPU load: avg 4% (peak 11%)

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