openai/gpt-oss-20b

gpt_oss · 20B · MXFP4

THINKING MODEL

Mac mini (Apple M4)

24 GB · macOS 26.3

Tested on March 19, 2026 · Submitted by cryptoepops

Top 25% Compare

Global Score

77 /100

Not Rec.

Hardware Fit

67/100

Quality

81/100

Get this model

Get it in LM Studio

Search and download models directly from the app

🤗

Find on HuggingFace

MLX versions & conversions

Hardware

Machine: Mac mini
CPU: Apple M4
Cores: 10 total (4 perf + 6 eff)
Frequency: 2.4 GHz
RAM: 24 GB LPDDR5
GPU: Apple M4
OS: macOS 26.3
Arch: arm64
Power Mode: balanced

Performance

Tokens/sec Tokens generated per second — higher is better: 30.5
Standard deviation Variation of speed across benchmark runs — lower means more consistent: ±1.0
First chunk latency Delay before the first streamed chunk arrives from runtime: 18 ms
Time to first token How fast the model starts responding — lower is better: 30.0 s
Load time Time to load the model into memory before inference: N/A
Memory usage RAM consumed during inference — percentage of total system memory: 15.8 GB (66%)
Total tokens Total number of tokens generated across all benchmark prompts: 1633
Thinking tokens (est.) Estimated internal reasoning tokens — models with "thinking" generate more tokens: ~954

Score breakdown

Speed

50/50

Time to first token

0/20

Memory

17/30

Quality

Reasoning

19/20

Coding

11/20

Instruction following

12/20

Structured output

15/15

Math

14/15

Multilingual

10/10

Category levels

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

Metadata

Spec version: 0.2.1
Runtime: LM Studio 0.4.7+4
Model format: MLX
Hardware profile: ENTRY
Result hash: db23b57b6f3224c815833d106078e6bfc3556d1d002b4bc7144e099190615692

Interpretation

Hardware fit: 67/100. Overall suitability: NOT RECOMMENDED (Global 77/100). Category profile: Reasoning: Strong, Coding: Adequate, Instruction Following: Adequate, Structured Output: Strong, Math: Strong, Multilingual: Strong.

Warnings

Token throughput is estimated from LM Studio output because native token stats were unavailable. Compare tok/s across backends cautiously.
Model memory footprint is estimated via LM Studio CLI rather than measured from a fresh load.

Disqualifiers

Time to first token too high: 30000ms (maximum: 21386ms for ENTRY profile)

Bench Environment

Power: AC CPU load: avg 21% (peak 23%)

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