aya:8b
OLLAMA GGUFcommand-r · 8.0B · F16
Mar 1, 2026 · Apple M4
qwen3.5:9b
OLLAMA GGUFqwen35 · 9.7B · Q4_K_M
Mar 7, 2026 · AMD Ryzen 5 5500
Global Score
66 vs 71
Hardware Fit
92 vs 56
Quality Score
55 vs 78
Hardware
aya:8b qwen3.5:9b
MachineMacBook AirSystem manufacturer
CPUApple M4AMD Ryzen 5 5500
Cores1012
RAM32 GB16 GB
GPUApple M4NVIDIA GeForce RTX 3060
OSmacOS 26.3Microsoft Windows 11 IoT Enterprise LTSC 10.0.26100
Archarm64x64
Power Modebalancedbalanced
Performance
aya:8b qwen3.5:9b
Tokens/sec20.05.3
First chunkN/A1624 ms
TTFT300 ms1.6 s
Load time0.9 s4.0 s
Memory usage6.0 GB7.9 GB
Memory %19%50%
HW Fit Score Breakdown
aya:8b
Speed
32/50
TTFT
30/20
Memory
30/30
qwen3.5:9b
Speed
11/50
TTFT
20/20
Memory
25/30
Quality
aya:8b
Reasoning
10/20
Coding
6/20
Instruction
13/20
Structured
15/15
Math
2/15
Multilingual
9/10
Reasoning: Adequate Coding: Weak Instruction Following: Adequate Structured Output: Strong Math: Poor Multilingual: Strong
qwen3.5:9b
Reasoning
14/20
Coding
17/20
Instruction
12/20
Structured
15/15
Math
10/15
Multilingual
10/10
Reasoning: Adequate Coding: Strong Instruction Following: Adequate 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