Compatibility check
Can I run Granite Speech 4.1 2b NAR on Apple M1 (8-core GPU)?
Yes — it runs.best quant:
Decode speed
34 tok/sest
Usable context
66K
Memory
16 GB
Fits in memory34 tok/sest · 66K ctx
Fits in memory29 tok/sest · 66K ctx
Fits in memory20 tok/sest · 66K ctx