WASTE runs the full 2.78-trillion-parameter Kimi K3 on a 64GB MacBook by streaming experts from NVMe.

sqliteai/waste (769 stars, created July 28) keeps only K3's 27.28GB trunk resident and streams the ~4% of experts activated per token off SSD, opening the undistilled model in 29.06GB of RAM. The counterintuitive result is the cache table: raising the expert cache from 17.32GB to 23.32GB lifts hit rate from 36.2% to 38.4% while throughput collapses eightfold, because a cache hit becomes a page fault. Storage bandwidth is binding: a cold token reads ~17GB of experts, and a USB enclosure at 0.94 GB/s versus internal NVMe at 12.78 GB/s is the difference between working and not. Layers validate against PyTorch to 3.6e-06 on final logits.