GPU

Graphics processing unit. For local AI the GPU is the component that accelerates inference: its video memory (VRAM) determines how large a model you can run. Apple's GPUs, built into Apple Silicon, run on unified memory instead of separate VRAM.