← Back to community benchmarks

Qwen3.5-4B

M3 Ultra (80c) · 512 GB · 4bit · 2026-03-19

Performance

1k

tokens

2,589

PP tok/s

151.1

TG tok/s

396

TTFT (ms)

4.0

Peak mem (GB)

Hardware

Chip M3 Ultra (80c)

Memory 512 GB

GPU Cores 80

Software

oMLX v0.2.18

macOS macOS 15.3

Context 1,024

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
1k	2,589	151.1	4.0 GB	current
4k	2,602	139.5	4.8 GB	view

Batching Results

Batch Size	TG tok/s	Speedup
1×	151.1	1.00×
2×	265.7	1.76×
4×	375.7	2.49×