Llama models run fine on the M2/M3 Macbooks thanks to llama.cpp/GGML.