DeepSeek-v4.1 Flash: Pushing the Limits of KV Cache Compression

A deep dive into the DeepSeek-V4.1 Flash technical report: CED, CSA2, HSI, Single-Pass mHC, Engram and FP4 KV Cache — how the KV cache was compressed to just 890 bytes per token.
109 points by mfiguiere 35 days ago zartbot.github.io 8 comments
Summary not yet generated for this article.

Comments (8 of ) ⌛

Loading more comments... ⌛
Scroll down to load more comments
No comments available.
Type to search HN and engineering blog articles
Press Escape to close  |  ↑↓ to navigate  |  Enter to open