huawei-csl/KVarN

KVarN is a native vLLM KV-cache quantization backend for your agents: 3-5x more context, throughput above FP16, and FP16-level accuracy. Calibration-free, one flag.

143 best points · 16 comments · 1 Hacker News post · 516 GitHub stars · View on GitHub

Hacker News discussions

• KVarN: Native vLLM backend for KV-cache quantization by Huawei (143 points, 16 comments, by theanonymousone, 2026-06-04)