Hi, M5+ Macs have a known optimization gap since we do not fully utilize the Metal 4 matmul operations in our kernels yet - so should be able to do much better on this specific comparison soon!