Compatibility check
Can I run Llama Guard 4 12B on Apple M3 Pro (18-core GPU)?
Yes — it runs.best quant:
Decode speed
14 tok/sest
Usable context
262K
Memory
36 GB
Fits in memory14 tok/sest · 262K ctx
Fits in memory12 tok/sest · 131K ctx
Fits in memory8 tok/sest · 131K ctx