qwen/qwen3.8-27b
LM-STUDIO MLXqwen3_5 · 27B · 4bit
Aug 18, 2026 · Apple M4 Max
Kimi-48B:48B
OLLAMA GGUFkimi-linear · 49.1B · unknown
Aug 19, 2026 · Intel Core™ i7-14700F
Global Score
87 vs 85
Hardware Fit
83 vs 83
Quality Score
88 vs 86
Hardware
qwen/qwen3.8-27b Kimi-48B:48B
MachineMacBook ProHP OMEN 35L VALORANT Limited Edition Gaming Desktop PC
CPUApple M4 MaxIntel Core™ i7-14700F
Cores1628
RAM128 GB64 GB
GPUApple M4 MaxNVIDIA GeForce RTX 5080
OSmacOS 26.5.2Microsoft Windows 11 Enterprise 10.0.26200
Archarm64x64
Power Modebalancedbalanced
Performance
qwen/qwen3.8-27b Kimi-48B:48B
Tokens/sec29.132.8
First chunk9 ms724 ms
TTFT650 ms724 ms
Load timeN/A20.9 s
Memory usage21.0 GB25.8 GB
Memory %16%41%
HW Fit Score Breakdown
qwen/qwen3.8-27b
Speed
33/50
TTFT
20/20
Memory
30/30
Kimi-48B:48B
Speed
36/50
TTFT
20/20
Memory
27/30
Quality
qwen/qwen3.8-27b
Reasoning
18/20
Coding
18/20
Instruction
16/20
Structured
15/15
Math
11/15
Multilingual
10/10
Reasoning: Strong Coding: Strong Instruction Following: Strong Structured Output: Strong Math: Adequate Multilingual: Strong
Kimi-48B:48B
Reasoning
16/20
Coding
17/20
Instruction
15/20
Structured
15/15
Math
13/15
Multilingual
10/10
Reasoning: Strong Coding: Strong 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