← Back to community benchmarks

Qwen2.5-7B-Instruct

M4 (10c) · 16 GB · 4bit · 2026-03-13

Performance

32k

tokens

162.6

PP tok/s

13.8

TG tok/s

201493

TTFT (ms)

6.2

Peak mem (GB)

Hardware

Chip M4 (10c)

Memory 16 GB

GPU Cores 10

Software

oMLX v0.2.6

macOS macOS 15.7.3

Context 32,768

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
4k	245.3	23.0	4.6 GB	view
8k	234.1	21.2	4.8 GB	view
16k	205.9	18.2	5.3 GB	view
32k	162.6	13.8	6.2 GB	current
64k	117.5	8.4	8.0 GB	view