← Back to community benchmarks

Llama-3.2-3B-Instruct

M1 Max (32c) · 32 GB · 4bit · 2026-03-10
Performance
4k
tokens
1,145
PP tok/s
101.6
TG tok/s
3577
TTFT (ms)
2.8
Peak mem (GB)
Hardware
Chip M1 Max (32c)
Memory 32 GB
GPU Cores 32
Software
oMLX v0.2.7
macOS macOS 26.0.1
Context 4,096
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,126 125.7 2.5 GB view
4k 1,145 101.6 2.8 GB current
Batching Results
Batch Size TG tok/s Speedup
125.7 1.00×
176.9 1.41×
192.1 1.53×