iOS and Metal Engineer @ Lapse, ex Prisma AI
Gemma 4 26B-A4B inference in ~2 GB of RAM on any M-series MacBook