qwen3.5:27b
OLLAMA GGUFqwen35 · 27.8B · Q4_K_M
Mar 5, 2026 · AMD RYZEN AI MAX+ 395
gemma4:12b
OLLAMA GGUFgemma4 · 11.9B · Q4_K_M
Aug 22, 2026 · Apple M1 Pro
Global Score
20 vs 63
Hardware Fit
37 vs 62
Quality Score
12 vs 63
Hardware
qwen3.5:27b gemma4:12b
MachineAZW GTR ProMacBook Pro
CPUAMD RYZEN AI MAX+ 395Apple M1 Pro
Cores3210
RAM125 GB16 GB
GPUAMD Radeon 8060SApple M1 Pro
OSUbuntu 24.04.4 LTSmacOS 26.6.2
Archx64arm64
Power Modelow-powerbalanced
Performance
qwen3.5:27b gemma4:12b
Tokens/sec10.615.3
First chunk358 ms1142 ms
TTFT13.7 s30.0 s
Load time10.8 s5.4 s
Memory usage39.7 GB7.6 GB
Memory %32%47%
HW Fit Score Breakdown
qwen3.5:27b
Speed
9/50
TTFT
0/20
Memory
28/30
gemma4:12b
Speed
36/50
TTFT
0/20
Memory
26/30
Quality
qwen3.5:27b
Reasoning
2/20
Coding
0/20
Instruction
2/20
Structured
0/15
Math
0/15
Multilingual
8/10
Reasoning: Poor Coding: Poor Instruction Following: Poor Structured Output: Poor Math: Poor Multilingual: Strong
gemma4:12b
Reasoning
10/20
Coding
10/20
Instruction
9/20
Structured
14/15
Math
10/15
Multilingual
10/10
Reasoning: Weak Coding: Adequate Instruction Following: Weak Structured Output: Strong Math: Adequate Multilingual: Strong
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