qwen3.5-2b-mlx

qwen3_5 · 2B · 8bit

MacBook Air (Apple M4)

32 GB · macOS 26.3

Tested on March 3, 2026

Top 86% Compare

Global Score

49 /100

Marginal

Hardware Fit

83/100

Quality

35/100

Get this model

🦙

Ollama

ollama pull qwen3.5-2b-mlx

View on Ollama Library

ollama.com/library/qwen3.5-2b-mlx

🤗

Find on HuggingFace

GGUF versions & quantizations

Hardware

Machine: MacBook Air
CPU: Apple M4
Cores: 10 total (4 perf + 6 eff)
Frequency: 2.4 GHz
RAM: 32 GB LPDDR5
GPU: Apple M4
OS: macOS 26.3
Arch: arm64
Power Mode: balanced

Performance

Tokens/sec Tokens generated per second — higher is better: 18.6
Standard deviation Variation of speed across benchmark runs — lower means more consistent: ±10.0
Time to first token How fast the model starts responding — lower is better: 2.9 s
Load time Time to load the model into memory before inference: 0.0 s
Memory usage RAM consumed during inference — percentage of total system memory: 0.0 GB (0%)
Total tokens Total number of tokens generated across all benchmark prompts: 725

Score breakdown

Speed

31/40

Time to first token

22/30

Memory

30/30

Quality

Reasoning

0/20

Coding

9/20

Instruction following

10/20

Structured output

4/15

Math

3/15

Multilingual

9/10

Category levels

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

Metadata

Spec version: 0.2.0
Runtime: Lm-studio unknown
Model format: GGUF
Hardware profile: BALANCED
Result hash: 4c0575399fffc431228fd1ddfdf6b75634b5367fae2d709c2bf5e4bc19ddae5f

Interpretation

Hardware fit: 83/100. Overall suitability: MARGINAL (Global 49/100). Category profile: Reasoning: Poor, Coding: Weak, Instruction Following: Adequate, Structured Output: Weak, Math: Poor, Multilingual: Strong.

Warnings

Token speed is unstable (stddev 10.0 tok/s, mean 18.6 tok/s) — may indicate thermal throttling or memory pressure.

Run yours now

npx metrillm@latest bench

Requires Node 20+ and Ollama or LM Studio running