← Back to community benchmarks

unsloth--Qwen3.6-35B-A3B-UD

M5 Pro (20c) · 64 GB · 4bit · 2026-07-18

Performance

32k

tokens

674.5

PP tok/s

31.1

TG tok/s

48581

TTFT (ms)

28.1

Peak mem (GB)

Hardware

Chip M5 Pro (20c)

Memory 64 GB

GPU Cores 20

Software

oMLX v0.5.1

macOS macOS 26.5.2

Context 32,768

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
1k	1,963	73.7	20.4 GB	view
4k	2,441	70.7	21.2 GB	view
8k	2,363	69.3	21.6 GB	view
16k	1,415	65.8	21.7 GB	view
32k	674.5	31.1	28.1 GB	current
64k	623.7	50.4	22.9 GB	view