Compatibility check
Can I run Gemma 4 26B A4B on Apple M3 Max (40-core GPU)?
Yes — it runs.best quant:
Decode speed
98 tok/sest
Usable context
262K
Memory
128 GB
Fits in memory98 tok/sest · 262K ctx
Fits in memory84 tok/sest · 262K ctx
Fits in memory56 tok/sest · 262K ctx