← Back to community benchmarks

Qwen3-4B-Instruct-2507

M3 Pro (18c) · 36 GB · 4bit · 2026-03-10

Performance

1k

tokens

591.7

PP tok/s

47.3

TG tok/s

1731

TTFT (ms)

2.9

Peak mem (GB)

Hardware

Chip M3 Pro (18c)

Memory 36 GB

GPU Cores 18

Software

oMLX v0.2.7

macOS macOS 26.3.1

Context 1,024

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
1k	591.7	47.3	2.9 GB	current
4k	553.0	38.9	3.4 GB	view

Batching Results

Batch Size	TG tok/s	Speedup
1×	47.3	1.00×
2×	88.4	1.87×
4×	126.6	2.68×