GPT-OSS-120B:120B

gptoss · 116.8B · MXFP4

THINKING MODEL

GMKtec NucBox EVO-T2S (Intel Core™ Ultra X7 358H)

64 GB · Microsoft Windows 11 Pro 10.0.26200

Tested on August 18, 2026 · Submitted by Haval

Top 68% Compare

Global Score

67 /100

Not Rec.

Hardware Fit

21/100

Quality

87/100

Get this model

🦙

Ollama

ollama pull GPT-OSS-120B:120B

View on Ollama Library

ollama.com/library/GPT-OSS-120B

Get it in LM Studio

Search and download models directly from the app

🤗

Find on HuggingFace

GGUF versions & quantizations

Hardware

Machine: GMKtec NucBox EVO-T2S
CPU: Intel Core™ Ultra X7 358H
Cores: 16 total (16 perf)
Frequency: 1.9 GHz
RAM: 64 GB LPDDR5
GPU: Intel(R) Arc(TM) B390 GPU
OS: Microsoft Windows 11 Pro 10.0.26200
Arch: x64
Power Mode: balanced

Performance

Tokens/sec Tokens generated per second — higher is better: 11.7
Standard deviation Variation of speed across benchmark runs — lower means more consistent: ±2.1
First chunk latency Delay before the first streamed chunk arrives from runtime: 4.3 s
Time to first token How fast the model starts responding — lower is better: 9.2 s
Load time Time to load the model into memory before inference: 84.1 s
Memory usage RAM consumed during inference — percentage of total system memory: 61.0 GB (96%)
Total tokens Total number of tokens generated across all benchmark prompts: 1513
Thinking tokens (est.) Estimated internal reasoning tokens — models with "thinking" generate more tokens: ~307

Score breakdown

Speed

15/50

Time to first token

4/20

Memory

2/30

Quality

Reasoning

18/20

Coding

19/20

Instruction following

10/20

Structured output

15/15

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 0.32.9
Model format: GGUF
Hardware profile: HIGH-END
Result hash: b8236f99896f39e10820965f0e62fbe52c9fad4af001d9bc49c84f4d20cf7309

Interpretation

Hardware fit: 21/100. Overall suitability: NOT RECOMMENDED (Global 67/100). Category profile: Reasoning: Strong, Coding: Strong, Instruction Following: Adequate, Structured Output: Strong, Math: Strong, Multilingual: Strong.

Warnings

Significant swap activity during benchmark (+9.7 GB). Model may exceed available RAM — results are severely degraded.

Disqualifiers

Memory usage critical: model delta +96%

Bench Environment

Swap delta: +9.7 GB CPU load: avg 51% (peak 52%)

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