DeepSeek-v4.1 Flash: Pushing the Limits of KV Cache Compression (zartbot.github.io) 2 points by mfiguiere 1h ago ↗ HN
[–] arikrahman 15m ago ↗ I am very impressed with the KV Cache Compression work as well as the prefix cacheing making queries converge on practically free.
1 comment
[ 0.19 ms ] story [ 3.7 ms ] thread