hf.co/empero-ai/Qwen3.8-4B-Distill-GGUF:Q4_K_M
OLLAMA GGUFqwen35 · 4.33B · unknown
Aug 28, 2026 · Apple M4
hf.co/unsloth/Qwen3.8-27B-GGUF:UD-IQ1_M
OLLAMA GGUFqwen35 · 27.3B · unknown
Aug 21, 2026 · AMD RYZEN AI MAX+ 395
Global Score
66 vs 68
Hardware Fit
100 vs 69
Quality Score
52 vs 67
Hardware
hf.co/empero-ai/Qw… hf.co/unsloth/Qwen…
MachineMacBook ProAZW GTR Pro
CPUApple M4AMD RYZEN AI MAX+ 395
Cores1032
RAM16 GB125 GB
GPUApple M4Radeon 8060S
OSmacOS 26.6.1Ubuntu 24.04.4 LTS
Archarm64x64
Power Modebalancedlow-power
Performance
hf.co/empero-ai/Qw… hf.co/unsloth/Qwen…
Tokens/sec26.521.1
First chunk447 ms571 ms
TTFT447 ms571 ms
Load time3.3 s4.9 s
Memory usage4.8 GB22.7 GB
Memory %30%18%
HW Fit Score Breakdown
hf.co/empero-ai/Qwen3.8…
Speed
50/50
TTFT
20/20
Memory
30/30
hf.co/unsloth/Qwen3.8-2…
Speed
20/50
TTFT
19/20
Memory
30/30
Quality
hf.co/empero-ai/Qwen3.8…
Reasoning
18/20
Coding
14/20
Instruction
4/20
Structured
1/15
Math
15/15
Multilingual
0/10
Reasoning: Strong Coding: Adequate Instruction Following: Poor Structured Output: Poor Math: Strong Multilingual: Poor
hf.co/unsloth/Qwen3.8-2…
Reasoning
12/20
Coding
10/20
Instruction
10/20
Structured
15/15
Math
10/15
Multilingual
10/10
Reasoning: Adequate Coding: Weak 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