qwen3.5:4b
OLLAMA GGUFqwen35 · 4.2B · Q4_K_M
Oct 12, 2026 · AMD Ryzen 5 5500
lfm2.5:8b
OLLAMA GGUFlfm2moe · 8.5B · Q4_K_M
Aug 24, 2026 · Apple M1 Pro
Global Score
81 vs 80
Hardware Fit
98 vs 95
Quality Score
74 vs 73
Hardware
qwen3.5:4b lfm2.5:8b
MachineMicrosoft WSLMacBook Pro
CPUAMD Ryzen 5 5500Apple M1 Pro
Cores1210
RAM8 GB16 GB
GPUIntegrated / UnknownApple M1 Pro
OSUbuntu 26.04.1 LTSmacOS 26.6.2
Archx64arm64
Power Modeunknownbalanced
Performance
qwen3.5:4b lfm2.5:8b
Tokens/sec93.798.5
First chunk249 ms167 ms
TTFT249 ms2.5 s
Load time10.1 s3.0 s
Memory usage2.9 GB6.0 GB
Memory %38%37%
HW Fit Score Breakdown
qwen3.5:4b
Speed
50/50
TTFT
20/20
Memory
28/30
lfm2.5:8b
Speed
50/50
TTFT
17/20
Memory
28/30
Quality
qwen3.5:4b
Reasoning
15/20
Coding
13/20
Instruction
15/20
Structured
15/15
Math
7/15
Multilingual
9/10
Reasoning: Strong Coding: Adequate Instruction Following: Strong Structured Output: Strong Math: Weak Multilingual: Strong
lfm2.5:8b
Reasoning
2/20
Coding
12/20
Instruction
19/20
Structured
15/15
Math
15/15
Multilingual
10/10
Reasoning: Poor Coding: Adequate Instruction Following: Strong Structured Output: Strong Math: Strong 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