← Back to community benchmarks

Qwen3.6-35B-A3B-PARO

M3 Ultra (60c) · 256 GB · 4bit · 2026-06-07

Performance

16k

tokens

1,966

PP tok/s

70.9

TG tok/s

8334

TTFT (ms)

20.3

Peak mem (GB)

Hardware

Chip M3 Ultra (60c)

Memory 256 GB

GPU Cores 60

Software

oMLX v0.4.2rc1

macOS macOS 26.5.1

Context 16,384

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
1k	1,558	73.7	18.6 GB	view
4k	2,121	73.1	19.3 GB	view
8k	2,108	70.6	19.7 GB	view
16k	1,966	70.9	20.3 GB	current
32k	1,680	65.5	21.6 GB	view
64k	1,280	56.1	24.3 GB	view