baidu/ernie-4.5-21b-a3b
LM-STUDIO GGUFernie4_5-moe · 21B · Q4_K_M
Nov 1, 2026 · AMD RYZEN AI MAX+ 395
qwen3.5:4b
OLLAMA GGUFqwen35 · 4.7B · Q4_K_M
Mar 2, 2026 · Apple M4
Global Score
86 vs 80
Hardware Fit
93 vs 88
Quality Score
83 vs 76
Hardware
baidu/ernie-4.5-21… qwen3.5:4b
MachineGMKtec EVO-X3MacBook Air
CPUAMD RYZEN AI MAX+ 395Apple M4
Cores3210
RAM120 GB32 GB
GPUAMD Radeon(TM) 8060S GraphicsApple M4
OSMicrosoft Windows 11 Pro 10.0.26300macOS 26.3
Archx64arm64
Power Modebalancedbalanced
Performance
baidu/ernie-4.5-21… qwen3.5:4b
Tokens/sec46.316.6
First chunk33 msN/A
TTFT383 ms369 ms
Load time49.7 s1.5 s
Memory usage32.5 GB6.5 GB
Memory %27%20%
HW Fit Score Breakdown
baidu/ernie-4.5-21b-a3b
Speed
43/50
TTFT
20/20
Memory
30/30
qwen3.5:4b
Speed
28/50
TTFT
30/20
Memory
30/30
Quality
baidu/ernie-4.5-21b-a3b
Reasoning
16/20
Coding
17/20
Instruction
15/20
Structured
14/15
Math
12/15
Multilingual
9/10
Reasoning: Strong Coding: Strong Instruction Following: Strong Structured Output: Strong Math: Strong Multilingual: Strong
qwen3.5:4b
Reasoning
16/20
Coding
14/20
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
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