Compatibility check
Can I run Gemma 4 26B A4B on Apple M5 Max (40-core GPU)?
Yes — it runs.best quant:
Decode speed
151 tok/sest
Usable context
262K
Memory
128 GB
Fits in memory151 tok/sest · 262K ctx
Fits in memory128 tok/sest · 262K ctx
Fits in memory86 tok/sest · 262K ctx