Compatibility check
Can I run Gemma 4 26B A4B on Apple M5 Max (32-core GPU)?
Yes — it runs.best quant:
Decode speed
113 tok/sest
Usable context
66K
Memory
64 GB
Fits in memory113 tok/sest · 66K ctx
Fits in memory96 tok/sest · 66K ctx
Fits in memory64 tok/sest · 66K ctx