Compatibility check
Can I run Llama 4 Scout on Apple M1 Ultra (64-core GPU)?
Yes — it runs.best quant:
Decode speed
9 tok/sest
Usable context
66K
Memory
128 GB
Fits in memory9 tok/sest · 66K ctx
Fits in memory7 tok/sest · 33K ctx
Spills to system RAM