Back to leaderboard

qwen3.5-4b-mlx

LM-STUDIO MLX

qwen3_5 · 4B · 4bit

Not Rec.

Mar 6, 2026 · Apple M4 Pro

deepseek-r1:70b

OLLAMA GGUF

llama · 70.6B · Q4_K_M

Not Rec.

Jun 27, 2026 · Cortex-X925

Global Score
62 vs 47
Hardware Fit
35 vs 51
Quality Score
73 vs 45

Hardware

qwen3.5-4b-mlx deepseek-r1:70b
MachineMac miniNVIDIA NVIDIA_DGX_Spark
CPUApple M4 ProCortex-X925
Cores1420
RAM64 GB122 GB
GPUApple M4 ProDevice 2e12
OSmacOS 15.7.4Ubuntu 24.04.4 LTS
Archarm64arm64
Power Modebalancedperformance

Performance

qwen3.5-4b-mlx deepseek-r1:70b
Tokens/sec0.04.7
First chunk5548 ms557 ms
TTFT5.5 s557 ms
Load timeN/A52.6 s
Memory usage2.9 GB49.5 GB
Memory %5%41%

HW Fit Score Breakdown

qwen3.5-4b-mlx

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

deepseek-r1:70b

Speed
4/50
TTFT
20/20
Memory
27/30

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

deepseek-r1:70b

Reasoning
3/20
Coding
1/20
Instruction
14/20
Structured
9/15
Math
8/15
Multilingual
10/10
Reasoning: Poor Coding: Poor Instruction Following: Adequate Structured Output: Adequate Math: Adequate Multilingual: Strong

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