glm-4.7-flash:latest
OLLAMA GGUFglm4moelite · 29.9B · Q4_K_M
Mar 22, 2026 · AMD RYZEN AI MAX+ 395
llama3:latest
OLLAMA GGUFllama · 8.0B · Q4_0
Mar 2, 2026 · Apple M4
Global Score
82 vs 71
Hardware Fit
94 vs 95
Quality Score
77 vs 60
Hardware
glm-4.7-flash:latest llama3:latest
MachineAZW GTR ProMacBook Air
CPUAMD RYZEN AI MAX+ 395Apple M4
Cores3210
RAM125 GB32 GB
GPURadeon 8060SApple M4
OSUbuntu 24.04.4 LTSmacOS 26.3
Archx64arm64
Power Modeperformancebalanced
Performance
glm-4.7-flash:latest llama3:latest
Tokens/sec47.622.3
First chunk200 msN/A
TTFT200 ms295 ms
Load time7.3 s0.8 s
Memory usage37.7 GB5.6 GB
Memory %30%18%
HW Fit Score Breakdown
glm-4.7-flash:latest
Speed
44/50
TTFT
20/20
Memory
30/30
llama3:latest
Speed
35/50
TTFT
30/20
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
llama3:latest
Reasoning
9/20
Coding
9/20
Instruction
15/20
Structured
13/15
Math
4/15
Multilingual
10/10
Reasoning: Weak Coding: Weak 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