gemma3n:latest

gemma3n · 6.9B · Q4_K_M

Supermicro AS -2015HS-TNR (AMD EPYC 9654P 96-Core Processor)

128 GB · Ubuntu 24.04.3 LTS

Tested on March 6, 2026

Top 16% Compare

Global Score

81 /100

Excellent

Hardware Fit

100/100

Quality

73/100

Get this model

🦙

Ollama

ollama pull gemma3n:latest

View on Ollama Library

ollama.com/library/gemma3n

Get it in LM Studio

Search and download models directly from the app

🤗

Find on HuggingFace

GGUF versions & quantizations

Hardware

Machine: Supermicro AS -2015HS-TNR
CPU: AMD EPYC 9654P 96-Core Processor
Cores: 64 total (-32 perf + 128 eff)
Frequency: 2.4 GHz
RAM: 128 GB
GPU: GH100 [H100L 94GB], ASPEED Graphics Family, GH100 [H100L 94GB]
OS: Ubuntu 24.04.3 LTS
Arch: x64
Power Mode: performance

Performance

Tokens/sec Tokens generated per second — higher is better: 108.7
Standard deviation Variation of speed across benchmark runs — lower means more consistent: ±3.1
First chunk latency Delay before the first streamed chunk arrives from runtime: 333 ms
Time to first token How fast the model starts responding — lower is better: 333 ms
Load time Time to load the model into memory before inference: 4.5 s
Memory usage RAM consumed during inference — percentage of total system memory: 8.6 GB (7%)
Total tokens Total number of tokens generated across all benchmark prompts: 1045

Score breakdown

Speed

50/50

Time to first token

20/20

Memory

30/30

Quality

Reasoning

14/20

Coding

14/20

Instruction following

14/20

Structured output

14/15

Math

7/15

Multilingual

10/10

Category levels

Reasoning: Adequate Coding: Adequate Instruction Following: Adequate Structured Output: Strong Math: Weak Multilingual: Strong

Metadata

Spec version: 0.2.1
Runtime: Ollama 0.17.5
Model format: GGUF
Hardware profile: HIGH-END
Result hash: feec5f950ade099b159e5524501d930f6b6fbfa2f7682627ba8bec54988dfaed

Interpretation

Hardware fit: 100/100. Overall suitability: EXCELLENT (Global 81/100). Category profile: Reasoning: Adequate, Coding: Adequate, Instruction Following: Adequate, Structured Output: Strong, Math: Weak, Multilingual: Strong.

Bench Environment

Thermal: nominal CPU load: avg 2% (peak 3%)

Run yours now

$ npm install -g metrillm@latest

$ metrillm

Requires Node 20+ and Ollama or LM Studio running

Or run without installing: npx metrillm@latest