Open post flangg @flangg@det.social · 4mo ago Replying to @finnvoorhees@mastodon.social a few optimizations later and it’s on par with MLX @finnvoorhees@mastodon.social Why is MLX much faster for prefill in the new benchmark?
mastodon 4.5.19 mastodon 4.5.19 mastodon 4.5.19