← Back to community benchmarks

Qwen3-30B-A3B

M3 Ultra (80c) · 512 GB · 6bit · 2026-07-14

Performance

16k

tokens

1,578

PP tok/s

62.5

TG tok/s

10381

TTFT (ms)

25.1

Peak mem (GB)

Hardware

Chip M3 Ultra (80c)

Memory 512 GB

GPU Cores 80

Software

oMLX v0.5.0

macOS macOS 26.5.1

Context 16,384

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
1k	2,135	89.5	22.8 GB	view
4k	2,359	84.4	23.4 GB	view
8k	2,027	75.5	24.1 GB	view
16k	1,578	62.5	25.1 GB	current
32k	1,083	50.1	27.4 GB	view
64k	661.3	36.6	31.9 GB	view
128k	371.9	23.6	41.0 GB	view
195k	257.8	17.1	53.7 GB	view