← Back to community benchmarks

Qwen3.5-35B-A3BMXFP4

M1 Max (32c) · 64 GB · 4bit · 2026-04-19

Performance

16k

tokens

691.0

PP tok/s

48.8

TG tok/s

23710

TTFT (ms)

22.8

Peak mem (GB)

Hardware

Chip M1 Max (32c)

Memory 64 GB

GPU Cores 32

Software

oMLX v0.3.6

macOS macOS 26.2

Context 16,384

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
1k	658.7	65.0	19.2 GB	view
4k	726.5	62.9	20.8 GB	view
8k	733.4	57.0	21.5 GB	view
16k	691.0	48.8	22.8 GB	current
32k	531.9	33.1	25.4 GB	view