6.4x faster than llama.cpp, 3.9x faster than MLX (basecompute.co) 4 points by lukasonedge 2mo ago ↗ HN
4 comments
[ 3.7 ms ] story [ 20.3 ms ] thread