nemotron-3-nano
LM-STUDIO GGUFMar 6, 2026 · Apple M4 Pro
qwen3.5-0.8b-mlx
LM-STUDIO MLXqwen3_5 · 0.8B · 4bit
Mar 3, 2026 · Apple M4
Global Score
93 vs 55
Hardware Fit
100 vs 87
Quality Score
90 vs 41
Hardware
nemotron-3-nano qwen3.5-0.8b-mlx
MachineMac miniMacBook Air
CPUApple M4 ProApple M4
Cores1410
RAM64 GB32 GB
GPUApple M4 ProApple M4
OSmacOS 15.7.4macOS 26.3
Archarm64arm64
Power Modebalancedbalanced
Performance
nemotron-3-nano qwen3.5-0.8b-mlx
Tokens/sec93.368.2
First chunk257 ms3268 ms
TTFT257 ms3.3 s
Load timeN/AN/A
Memory usage0.1 GB0.6 GB
Memory %0%2%
HW Fit Score Breakdown
nemotron-3-nano
Speed
50/50
TTFT
20/20
Memory
30/30
qwen3.5-0.8b-mlx
Speed
50/50
TTFT
13/20
Memory
24/30
Quality
nemotron-3-nano
Reasoning
19/20
Coding
18/20
Instruction
14/20
Structured
15/15
Math
14/15
Multilingual
10/10
Reasoning: Strong Coding: Strong Instruction Following: Adequate Structured Output: Strong Math: Strong Multilingual: Strong
qwen3.5-0.8b-mlx
Reasoning
8/20
Coding
1/20
Instruction
12/20
Structured
12/15
Math
3/15
Multilingual
5/10
Reasoning: Weak Coding: Poor Instruction Following: Adequate Structured Output: Strong Math: Poor Multilingual: Adequate
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