baidu/ernie-4.5-21b-a3b
LM-STUDIO GGUFernie4_5-moe · 21B · Q4_K_M
Nov 1, 2026 · AMD RYZEN AI MAX+ 395
huihui_ai/Qwen3.8-abliterated:27b
OLLAMA GGUFqwen35 · 27.3B · Q4_K_M
Aug 25, 2026 · AMD RYZEN AI MAX+ 395
Global Score
86 vs 80
Hardware Fit
93 vs 65
Quality Score
83 vs 87
Hardware
baidu/ernie-4.5-21… huihui_ai/Qwen3.8-…
MachineGMKtec EVO-X3AZW GTR Pro
CPUAMD RYZEN AI MAX+ 395AMD RYZEN AI MAX+ 395
Cores3232
RAM120 GB125 GB
GPUAMD Radeon(TM) 8060S GraphicsRadeon 8060S
OSMicrosoft Windows 11 Pro 10.0.26300Ubuntu 24.04.4 LTS
Archx64x64
Power Modebalancedlow-power
Performance
baidu/ernie-4.5-21… huihui_ai/Qwen3.8-…
Tokens/sec46.317.5
First chunk33 ms597 ms
TTFT383 ms597 ms
Load time49.7 s7.1 s
Memory usage32.5 GB17.4 GB
Memory %27%14%
HW Fit Score Breakdown
baidu/ernie-4.5-21b-a3b
Speed
43/50
TTFT
20/20
Memory
30/30
huihui_ai/Qwen3.8-ablit…
Speed
16/50
TTFT
19/20
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
huihui_ai/Qwen3.8-ablit…
Reasoning
17/20
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
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