drewbitt on Matmul() using PyTorch's MPS back end is faster than Apple's MLX · Hackerly

Comment on Matmul() using PyTorch's MPS back end is faster than Apple's MLX

I have noticed the same thing with PyTorch and MLX.