← Back to community benchmarks

Llama-3.2-3B-Instruct

M4 Pro (20c) · 48 GB · 4bit · 2026-03-12

Performance

1k

tokens

1,036

PP tok/s

104.5

TG tok/s

989

TTFT (ms)

2.5

Peak mem (GB)

Hardware

Chip M4 Pro (20c)

Memory 48 GB

GPU Cores 20

Software

oMLX v0.2.9

macOS macOS 26.3.1

Context 1,024

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
1k	1,036	104.5	2.5 GB	current
4k	1,022	86.8	2.8 GB	view

Batching Results

Batch Size	TG tok/s	Speedup
1×	104.5	1.00×
2×	186.9	1.79×
4×	215.5	2.06×