← Back to community benchmarks

Qwen3.5-9B

M5 (10c) · 32 GB · 8bit · 2026-07-19

Performance

32k

tokens

569.2

PP tok/s

11.4

TG tok/s

57569

TTFT (ms)

13.1

Peak mem (GB)

Hardware

Chip M5 (10c)

Memory 32 GB

GPU Cores 10

Software

oMLX v0.5.2.dev1

macOS macOS 26.5.2

Context 32,768

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
16k	659.9	12.3	11.6 GB	view
32k	569.2	11.4	13.1 GB	current
64k	447.8	9.9	16.2 GB	view