← Back to community benchmarks

Qwen3.6-27B-oQ4-mtp

M3 Ultra (80c) · 512 GB · 4bit · 2026-07-12

Performance

64k

tokens

345.6

PP tok/s

26.2

TG tok/s

189644

TTFT (ms)

21.7

Peak mem (GB)

Hardware

Chip M3 Ultra (80c)

Memory 512 GB

GPU Cores 80

Software

oMLX v0.5.1

macOS macOS 26.4

Context 65,536

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
8k	432.5	32.8	17.9 GB	view
64k	345.6	26.2	21.7 GB	current
128k	275.4	21.6	27.1 GB	view