qwen3:4b
OLLAMA GGUFqwen3 · 4.0B · Q4_K_M
Jun 26, 2026 · Cortex-A78AE
qwen3:4b
OLLAMA GGUFqwen3 · 4.0B · Q4_K_M
Jun 26, 2026 · Cortex-A78AE
Global Score
32 vs 29
Hardware Fit
98 vs 86
Quality Score
4 vs 4
Hardware
qwen3:4b qwen3:4b
MachineNVIDIA NVIDIA Jetson Orin Nano Engineering Reference Developer Kit SuperNVIDIA NVIDIA Jetson Orin Nano Engineering Reference Developer Kit Super
CPUCortex-A78AECortex-A78AE
Cores66
RAM7 GB7 GB
GPUIntegrated / UnknownIntegrated / Unknown
OSUbuntu 22.04.5 LTSUbuntu 22.04.5 LTS
Archarm64arm64
Power Modebalancedbalanced
Performance
qwen3:4b qwen3:4b
Tokens/sec17.917.9
First chunk950 ms1010 ms
TTFT950 ms6.4 s
Load time9.7 s9.5 s
Memory usage3.0 GB3.0 GB
Memory %40%40%
HW Fit Score Breakdown
qwen3:4b
Speed
50/50
TTFT
20/20
Memory
28/30
qwen3:4b
Speed
50/50
TTFT
8/20
Memory
28/30
Quality
qwen3:4b
Reasoning
4/20
Coding
0/20
Instruction
0/20
Structured
0/15
Math
0/15
Multilingual
0/10
Reasoning: Poor Coding: Poor Instruction Following: Poor Structured Output: Poor Math: Poor Multilingual: Poor
qwen3:4b
Reasoning
4/20
Coding
0/20
Instruction
0/20
Structured
0/15
Math
0/15
Multilingual
0/10
Reasoning: Poor Coding: Poor Instruction Following: Poor Structured Output: Poor Math: Poor Multilingual: Poor
Run yours and compare
$
npm install -g metrillm@latest$
metrillmRequires Node 20+ and Ollama or LM Studio running
Or run without installing: npx metrillm@latest