← Back to community benchmarks

Llama-3.2-3B-Instruct

M2 Max (38c) · 32 GB · 4bit · 2026-03-12
Performance
4k
tokens
1,289
PP tok/s
103.8
TG tok/s
3177
TTFT (ms)
2.8
Peak mem (GB)
Hardware
Chip M2 Max (38c)
Memory 32 GB
GPU Cores 38
Software
oMLX v0.2.8
macOS macOS 26.3
Context 4,096
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,222 127.2 2.5 GB view
4k 1,289 103.8 2.8 GB current
Batching Results
Batch Size TG tok/s Speedup
1× 127.2 1.00×
2× 198.5 1.56×
4× 228.8 1.80×