← Back to community benchmarks

Llama-3.2-3B-Instruct

M1 (8c) · 16 GB · 4bit · 2026-03-12
Performance
4k
tokens
292.3
PP tok/s
23.8
TG tok/s
14019
TTFT (ms)
2.6
Peak mem (GB)
Hardware
Chip M1 (8c)
Memory 16 GB
GPU Cores 8
Software
oMLX v0.2.9
macOS macOS 26.3.1
Context 4,096
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 312.3 28.6 2.3 GB view
4k 292.3 23.8 2.6 GB current
Batching Results
Batch Size TG tok/s Speedup
1× 28.6 1.00×
2× 52.5 1.84×
4× 57.0 1.99×
8× 58.1 2.03×