qwen3.5-27b-claude-4.6-opus-distilled-mlx

qwen3_5 · 27B · 4bit

THINKING MODEL

Mac mini (Apple M4)

24 GB · macOS 26.3

Tested on March 10, 2026

Top 95% Compare

Global Score

34 /100

Not Rec.

Hardware Fit

21/100

Quality

39/100

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: 6.1
Standard deviation Variation of speed across benchmark runs — lower means more consistent: ±0.6
First chunk latency Delay before the first streamed chunk arrives from runtime: 7 ms
Time to first token How fast the model starts responding — lower is better: 22.9 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: 19.8 GB (82%)
Total tokens Total number of tokens generated across all benchmark prompts: 1729
Thinking tokens (est.) Estimated internal reasoning tokens — models with "thinking" generate more tokens: ~553

Score breakdown

Speed

12/50

Time to first token

0/20

Memory

9/30

Quality

Reasoning

1/20

Coding

14/20

Instruction following

5/20

Structured output

8/15

Math

4/15

Multilingual

7/10

Category levels

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

Metadata

Spec version: 0.2.1
Runtime: LM Studio 0.4.6+1
Model format: MLX
Hardware profile: ENTRY
Result hash: 9d2f7dbabebf5656b57cbc30df836ebc3a2b33c29ad4fa84231ce3aa3fc6bf52

Interpretation

Hardware fit: 21/100. Overall suitability: NOT RECOMMENDED (Global 34/100). Category profile: Reasoning: Poor, Coding: Adequate, Instruction Following: Weak, Structured Output: Adequate, Math: Poor, Multilingual: Adequate.

Warnings

Model memory footprint is estimated via LM Studio CLI rather than measured from a fresh load.

Disqualifiers

Time to first token too high: 22856ms (maximum: 21386ms for ENTRY profile)

Bench Environment

Power: AC CPU load: avg 9% (peak 14%)

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