The M5 macs just dropped.
It looks like we have a neural engine in every core so the base M5 might have the equivalent prefill speed of the maxed m4 max 40 core gpu.
10 core GPU m4 = 1x
40 core GPU m4 = 4x
If the new baseline cheapest mac all can run this fast, it unlocks a lot of local AI software options like realtime transcription or realtime VLM work.
Yuji Tomita 