Hacker News new | ask | show | jobs
DeepSeek-V4 KV Cache Explained: Why 1M Context Uses Less VRAM (knightli.com)
3 points by vinhnx 18 days ago