gpt-oss-safeguard-20b-mlx
LM-STUDIO MLXgpt_oss · 20B · MXFP4
Mar 12, 2026 · Apple M4
mlx-community/Nanbeige4.1-3B-8bit
LM-STUDIO GGUFMar 5, 2026 · Apple M4
Global Score
87 vs 53
Hardware Fit
82 vs 99
Quality Score
89 vs 33
Hardware
gpt-oss-safeguard-… mlx-community/Nanb…
MachineMac miniMacBook Air
CPUApple M4Apple M4
Cores1010
RAM24 GB32 GB
GPUApple M4Apple M4
OSmacOS 26.3macOS 26.3
Archarm64arm64
Power Modebalancedbalanced
Performance
gpt-oss-safeguard-… mlx-community/Nanb…
Tokens/sec41.425.4
First chunk8 ms266 ms
TTFT3.6 s266 ms
Load timeN/AN/A
Memory usage14.5 GB0.0 GB
Memory %61%0%
HW Fit Score Breakdown
gpt-oss-safeguard-20b-mlx
Speed
50/50
TTFT
12/20
Memory
20/30
mlx-community/Nanbeige4…
Speed
49/50
TTFT
20/20
Memory
30/30
Quality
gpt-oss-safeguard-20b-mlx
Reasoning
19/20
Coding
12/20
Instruction
19/20
Structured
15/15
Math
15/15
Multilingual
9/10
Reasoning: Strong Coding: Adequate Instruction Following: Strong Structured Output: Strong Math: Strong Multilingual: Strong
mlx-community/Nanbeige4…
Reasoning
10/20
Coding
7/20
Instruction
1/20
Structured
1/15
Math
11/15
Multilingual
3/10
Reasoning: Adequate Coding: Weak Instruction Following: Poor Structured Output: Poor Math: Adequate Multilingual: Weak
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