← Back to community benchmarks

agentscope-ai:QwenPaw-Flash-9B-oQ4-fp16-mtp

M4 (10c) · 16 GB · 4bit · 2026-05-15

Performance

4k

tokens

225.6

PP tok/s

20.0

TG tok/s

18156

TTFT (ms)

6.4

Peak mem (GB)

Hardware

Chip M4 (10c)

Memory 16 GB

GPU Cores 10

Software

oMLX v0.3.9.dev1

macOS macOS 26.5

Context 4,096

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
1k	224.0	19.9	5.8 GB	view
4k	225.6	20.0	6.4 GB	current

Batching Results

Batch Size	TG tok/s	Speedup
1×	19.9	1.00×
2×	39.3	1.97×
4×	81.0	4.07×