← Back to community benchmarks

Qwen3.5-9b-Sushi-Coder-RL

M5 (10c) · 32 GB · 4bit · 2026-05-04

Performance

32k

tokens

1,546

PP tok/s

16.7

TG tok/s

21198

TTFT (ms)

8.6

Peak mem (GB)

Hardware

Chip M5 (10c)

Memory 32 GB

GPU Cores 10

Software

oMLX v0.3.8

macOS macOS 26.4.1

Context 32,768

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
1k	642.8	101.3	8.1 GB	view
4k	454.3	95.1	9.4 GB	view
8k	386.5	89.2	9.7 GB	view
16k	1,362	20.4	7.8 GB	view
32k	1,546	16.7	8.6 GB	current
64k	1,244	17.9	9.7 GB	view
128k	856.5	14.8	12.0 GB	view