Back to leaderboard

qwen3.5-4b-mlx

LM-STUDIO MLX

qwen3_5 · 4B · 4bit

Not Rec.

Mar 6, 2026 · Apple M4 Pro

google/gemma-4-e4b

LM-STUDIO GGUF

gemma4 · Q4_K_M

Good

Jul 7, 2026 · AMD Ryzen 7 5800H

Global Score
62 vs 62
Hardware Fit
35 vs 77
Quality Score
73 vs 55

Hardware

qwen3.5-4b-mlx google/gemma-4-e4b
MachineMac miniLENOVO 82JQ
CPUApple M4 ProAMD Ryzen 7 5800H
Cores1416
RAM64 GB16 GB
GPUApple M4 ProNVIDIA GeForce RTX 3060 Laptop GPU
OSmacOS 15.7.4Microsoft Windows 11 家庭版 中文版 10.0.26100
Archarm64x64
Power Modebalancedbalanced

Performance

qwen3.5-4b-mlx google/gemma-4-e4b
Tokens/sec0.027.6
First chunk5548 ms23 ms
TTFT5.5 s480 ms
Load timeN/AN/A
Memory usage2.9 GB13.2 GB
Memory %5%83%

HW Fit Score Breakdown

qwen3.5-4b-mlx

Speed
0/50
TTFT
6/20
Memory
29/30

google/gemma-4-e4b

Speed
48/50
TTFT
20/20
Memory
9/30

Quality

qwen3.5-4b-mlx

Reasoning
17/20
Coding
17/20
Instruction
9/20
Structured
8/15
Math
13/15
Multilingual
9/10
Reasoning: Strong Coding: Strong Instruction Following: Weak Structured Output: Adequate Math: Strong Multilingual: Strong

google/gemma-4-e4b

Reasoning
18/20
Coding
17/20
Instruction
5/20
Structured
3/15
Math
11/15
Multilingual
1/10
Reasoning: Strong Coding: Strong Instruction Following: Weak Structured Output: Poor Math: Strong Multilingual: Poor

Run yours and compare

$ npm install -g metrillm@latest
$ metrillm

Requires Node 20+ and Ollama or LM Studio running

Or run without installing: npx metrillm@latest