gpt-oss:120b

gptoss · 116.8B · MXFP4

THINKING MODEL

NVIDIA NVIDIA_DGX_Spark (Cortex-X925)

122 GB · Ubuntu 24.04.5 LTS

Tested on October 2, 2026 · Submitted by Tram

Top 30% Compare

Global Score

83 /100

Excellent

Hardware Fit

64/100

Quality

91/100

Get this model

🦙

Ollama

ollama pull gpt-oss:120b

View on Ollama Library

ollama.com/library/gpt-oss

Get it in LM Studio

Search and download models directly from the app

🤗

Find on HuggingFace

GGUF versions & quantizations

Hardware

Machine: NVIDIA NVIDIA_DGX_Spark
CPU: Cortex-X925
Cores: 20 total (20 perf)
Frequency: 3.35 GHz
RAM: 122 GB
GPU: Device 2e12
OS: Ubuntu 24.04.5 LTS
Arch: arm64
Power Mode: performance

Performance

Tokens/sec Tokens generated per second — higher is better: 28.6
Standard deviation Variation of speed across benchmark runs — lower means more consistent: ±11.8
First chunk latency Delay before the first streamed chunk arrives from runtime: 422 ms
Time to first token How fast the model starts responding — lower is better: 2.5 s
Load time Time to load the model into memory before inference: 9.9 s
Memory usage RAM consumed during inference — percentage of total system memory: 60.1 GB (49%)
Total tokens Total number of tokens generated across all benchmark prompts: 1483
Thinking tokens (est.) Estimated internal reasoning tokens — models with "thinking" generate more tokens: ~153

Score breakdown

Speed

30/50

Time to first token

9/20

Memory

25/30

Quality

Reasoning

19/20

Coding

19/20

Instruction following

14/20

Structured output

14/15

Math

15/15

Multilingual

10/10

Category levels

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

Metadata

Spec version: 0.2.1
Runtime: Ollama 0.35.0
Model format: GGUF
Hardware profile: HIGH-END
Result hash: 4a2942237f38c07cedacf3ecb0db937c0e105fcb6a5a09abf0b250c7c80770e0

Interpretation

Hardware fit: 64/100. Overall suitability: EXCELLENT (Global 83/100). Category profile: Reasoning: Strong, Coding: Strong, Instruction Following: Adequate, Structured Output: Strong, Math: Strong, Multilingual: Strong.

Warnings

Token speed is unstable (stddev 11.8 tok/s, mean 28.6 tok/s) — may indicate thermal throttling or memory pressure.

Bench Environment

Thermal: nominal CPU load: avg 21% (peak 28%)

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