qwen3:8b
OLLAMA GGUFqwen3 · 8.2B · Q4_K_M
Mar 5, 2026 · Intel Core™ Ultra 9 285K
mlx-community/Nanbeige4.1-3B-8bit
LM-STUDIO GGUFMar 5, 2026 · Apple M4
Global Score
84 vs 53
Hardware Fit
97 vs 99
Quality Score
79 vs 33
Hardware
qwen3:8b mlx-community/Nanb…
MachineASUSMacBook Air
CPUIntel Core™ Ultra 9 285KApple M4
Cores2410
RAM47 GB32 GB
GPUNVIDIA GeForce RTX 5090, Intel(R) GraphicsApple M4
OSMicrosoft Windows 11 Pro 10.0.26200macOS 26.3
Archx64arm64
Power Modebalancedbalanced
Performance
qwen3:8b mlx-community/Nanb…
Tokens/sec208.025.4
First chunk82 ms266 ms
TTFT1.3 s266 ms
Load time1.5 sN/A
Memory usage9.6 GB0.0 GB
Memory %20%0%
HW Fit Score Breakdown
qwen3:8b
Speed
50/50
TTFT
18/20
Memory
29/30
mlx-community/Nanbeige4…
Speed
49/50
TTFT
20/20
Memory
30/30
Quality
qwen3:8b
Reasoning
18/20
Coding
10/20
Instruction
14/20
Structured
15/15
Math
12/15
Multilingual
10/10
Reasoning: Strong Coding: Weak Instruction Following: Adequate Structured Output: Strong Math: Strong Multilingual: Strong
mlx-community/Nanbeige4…
Reasoning
10/20
Coding
7/20
Instruction
1/20
Structured
1/15
Math
11/15
Multilingual
3/10
Reasoning: Adequate Coding: Weak Instruction Following: Poor Structured Output: Poor Math: Adequate Multilingual: Weak
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