Matmul() using PyTorch's MPS back end is faster than Apple's MLX kevinmartinjose.com 1 points by lonesword 8 months ago · 1 comment Reader PiP Save drewbitt 8 months ago I have noticed the same thing with PyTorch and MLX.