gemma-3-270m-it-qat-mlx
LM-STUDIO MLXgemma3_text · 270M · 4bit
Mar 4, 2026 · Apple M4
llama3:latest
OLLAMA GGUFllama · 8.0B · Q4_0
Mar 2, 2026 · Apple M4
Global Score
43 vs 71
Hardware Fit
99 vs 95
Quality Score
19 vs 60
Hardware
gemma-3-270m-it-qa… llama3:latest
MachineMacBook AirMacBook Air
CPUApple M4Apple M4
Cores1010
RAM32 GB32 GB
GPUApple M4Apple M4
OSmacOS 26.3macOS 26.3
Archarm64arm64
Power Modebalancedbalanced
Performance
gemma-3-270m-it-qa… llama3:latest
Tokens/sec344.222.3
First chunk173 msN/A
TTFT173 ms295 ms
Load timeN/A0.8 s
Memory usage0.3 GB5.6 GB
Memory %1%18%
HW Fit Score Breakdown
gemma-3-270m-it-qat-mlx
Speed
50/50
TTFT
20/20
Memory
29/30
llama3:latest
Speed
35/50
TTFT
30/20
Memory
30/30
Quality
gemma-3-270m-it-qat-mlx
Reasoning
4/20
Coding
0/20
Instruction
9/20
Structured
3/15
Math
1/15
Multilingual
2/10
Reasoning: Poor Coding: Poor Instruction Following: Weak Structured Output: Poor Math: Poor Multilingual: Poor
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