glm-4.7-flash:latest
OLLAMA GGUFglm4moelite · 29.9B · Q4_K_M
Mar 22, 2026 · AMD RYZEN AI MAX+ 395
starcoder2:3b
OLLAMA GGUFstarcoder2 · 3B · Q4_0
Jun 26, 2026 · Cortex-A78AE
Global Score
82 vs 35
Hardware Fit
94 vs 100
Quality Score
77 vs 7
Hardware
glm-4.7-flash:latest starcoder2:3b
MachineAZW GTR ProNVIDIA NVIDIA Jetson Orin Nano Engineering Reference Developer Kit Super
CPUAMD RYZEN AI MAX+ 395Cortex-A78AE
Cores326
RAM125 GB7 GB
GPURadeon 8060SIntegrated / Unknown
OSUbuntu 24.04.4 LTSUbuntu 22.04.5 LTS
Archx64arm64
Power Modeperformancebalanced
Performance
glm-4.7-flash:latest starcoder2:3b
Tokens/sec47.631.5
First chunk200 ms392 ms
TTFT200 ms392 ms
Load time7.3 s4.8 s
Memory usage37.7 GB1.8 GB
Memory %30%24%
HW Fit Score Breakdown
glm-4.7-flash:latest
Speed
44/50
TTFT
20/20
Memory
30/30
starcoder2:3b
Speed
50/50
TTFT
20/20
Memory
30/30
Quality
glm-4.7-flash:latest
Reasoning
13/20
Coding
17/20
Instruction
12/20
Structured
15/15
Math
10/15
Multilingual
10/10
Reasoning: Adequate Coding: Strong Instruction Following: Adequate Structured Output: Strong Math: Adequate Multilingual: Strong
starcoder2:3b
Reasoning
0/20
Coding
3/20
Instruction
1/20
Structured
0/15
Math
1/15
Multilingual
2/10
Reasoning: Poor Coding: Poor Instruction Following: Poor Structured Output: Poor Math: Poor Multilingual: Poor
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