qwen3:4b
OLLAMA GGUFqwen3 · 4.0B · Q4_K_M
Jun 26, 2026 · Cortex-A78AE
qwen3.6:27b
OLLAMA GGUFqwen35 · 27.8B · Q4_K_M
Jun 25, 2026 · Apple M4 Pro
Global Score
29 vs 20
Hardware Fit
86 vs 34
Quality Score
4 vs 14
Hardware
qwen3:4b qwen3.6:27b
MachineNVIDIA NVIDIA Jetson Orin Nano Engineering Reference Developer Kit SuperMac mini
CPUCortex-A78AEApple M4 Pro
Cores612
RAM7 GB24 GB
GPUIntegrated / UnknownApple M4 Pro
OSUbuntu 22.04.5 LTSmacOS 15.6.1
Archarm64arm64
Power Modebalancedbalanced
Performance
qwen3:4b qwen3.6:27b
Tokens/sec17.910.6
First chunk1010 ms1344 ms
TTFT6.4 s30.0 s
Load time9.5 s0.2 s
Memory usage3.0 GB18.1 GB
Memory %40%75%
HW Fit Score Breakdown
qwen3:4b
Speed
50/50
TTFT
8/20
Memory
28/30
qwen3.6:27b
Speed
22/50
TTFT
0/20
Memory
12/30
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
qwen3.6:27b
Reasoning
2/20
Coding
1/20
Instruction
2/20
Structured
1/15
Math
2/15
Multilingual
6/10
Reasoning: Poor Coding: Poor Instruction Following: Poor Structured Output: Poor Math: Poor Multilingual: Adequate
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