nemotron-3-nano
LM-STUDIO GGUFMar 6, 2026 · Apple M4 Pro
qwen2.5-math-1.5b-instruct
LM-STUDIO MLXqwen2 · 1.5B · 4bit
Mar 4, 2026 · Apple M4
Global Score
93 vs 45
Hardware Fit
100 vs 99
Quality Score
90 vs 22
Hardware
nemotron-3-nano qwen2.5-math-1.5b-…
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 qwen2.5-math-1.5b-…
Tokens/sec93.391.7
First chunk257 ms205 ms
TTFT257 ms205 ms
Load timeN/AN/A
Memory usage0.1 GB0.8 GB
Memory %0%3%
HW Fit Score Breakdown
nemotron-3-nano
Speed
50/50
TTFT
20/20
Memory
30/30
qwen2.5-math-1.5b-instr…
Speed
50/50
TTFT
20/20
Memory
29/30
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
qwen2.5-math-1.5b-instr…
Reasoning
2/20
Coding
0/20
Instruction
3/20
Structured
1/15
Math
14/15
Multilingual
2/10
Reasoning: Poor Coding: Poor Instruction Following: Poor Structured Output: Poor Math: Strong 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