Compatibility check
Can I run Mistral Small 3 on Apple M4 Max (40-core GPU)?
Yes — it runs.best quant:
Decode speed
27 tok/sest
Usable context
33K
Memory
128 GB
Fits in memory27 tok/sest · 33K ctx
Fits in memory23 tok/sest · 33K ctx
Fits in memory15 tok/sest · 33K ctx