Compatibility check
Can I run Mixtral 8x22B Instruct on Apple M2 Ultra (76-core GPU)?
Yes — it runs.best quant:
Decode speed
20 tok/sest
Usable context
66K
Memory
192 GB
Fits in memory20 tok/sest · 66K ctx
Fits in memory17 tok/sest · 66K ctx
Spills to system RAM
