Compatibility check
Can I run Llama 3 8B Lunaris on Apple M3 Ultra (80-core GPU)?
Yes — it runs.best quant:
Decode speed
118 tok/sest
Usable context
8K
Memory
512 GB
Fits in memory118 tok/sest · 8K ctx
Fits in memory101 tok/sest · 8K ctx
Fits in memory67 tok/sest · 8K ctx