← Back to community benchmarks

Llama-3.2-3B-Instruct

M4 (10c) · 32 GB · 8bit · 2026-03-11
Performance
4k
tokens
513.6
PP tok/s
25.4
TG tok/s
7978
TTFT (ms)
4.1
Peak mem (GB)
Hardware
Chip M4 (10c)
Memory 32 GB
GPU Cores 10
Software
oMLX v0.2.7
macOS macOS 26.3.1
Context 4,096
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 548.9 28.5 3.7 GB view
4k 513.6 25.4 4.1 GB current
8k 468.5 22.6 4.5 GB view
16k 388.6 18.3 5.4 GB view
32k 282.2 13.5 7.3 GB view
64k 181.5 8.7 11.0 GB view
Batching Results
Batch Size TG tok/s Speedup
28.5 1.00×
54.8 1.92×
85.2 2.99×