# KVarN: Native vLLM backend for KV-cache quantization by Huawei

**URL:** <https://forums.developer.nvidia.com/t/kvarn-native-vllm-backend-for-kv-cache-quantization-by-huawei/372333>\
**Category:** DGX Spark / GB10 Projects\
**Tags:** agentic-ai\
**Created:** [June 4, 2026, 6:42pm UTC](https://forums.developer.nvidia.com/t/kvarn-native-vllm-backend-for-kv-cache-quantization-by-huawei/372333 "2026-06-04T18:42:36Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![adg1](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@adg1](https://forums.developer.nvidia.com/u/adg1)\
**Post date:** [June 4, 2026, 6:42pm UTC](https://forums.developer.nvidia.com/t/kvarn-native-vllm-backend-for-kv-cache-quantization-by-huawei/372333/1 "2026-06-04T18:42:36Z")

</div>

> ⚡️ **Built for agentic and long-context workloads.**

> 💡 KVarN delivers **3-5x more KV-cache capacity** and **up to ~1.3x the throughput** of FP16, so you fit far longer contexts and serve more concurrent requests, with **FP16-level accuracy**.

> 🔌 **Calibration-free, plug-and-play with vLLM.** A native vLLM attention backend: add one flag, no model changes, no calibration.

> 🥊 **Up to ~2.4× TurboQuant throughput** , same capacity, **higher accuracy**.

> **[GitHub - huawei-csl/KVarN: KVarN is a native vLLM KV-cache quantization...](https://github.com/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.

---

<div class="post-metadata">

**Author:** ![JW2026](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/jw2026/32/523471_2.png) [@JW2026](https://forums.developer.nvidia.com/u/JW2026)\
**Post date:** [June 8, 2026, 10:11am UTC](https://forums.developer.nvidia.com/t/kvarn-native-vllm-backend-for-kv-cache-quantization-by-huawei/372333/2 "2026-06-08T10:11:44Z")

</div>

> [@adg1](#):
>
> no model changes

Have you tried ? It doesn’t work with QWEN 3.6 , so the whole “no model changes” isn’t exactly accurate.

---

<div class="post-metadata">

**Author:** ![adg1](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@adg1](https://forums.developer.nvidia.com/u/adg1)\
**Post date:** [June 8, 2026, 11:03am UTC](https://forums.developer.nvidia.com/t/kvarn-native-vllm-backend-for-kv-cache-quantization-by-huawei/372333/3 "2026-06-08T11:03:33Z")

</div>

It doesn’t surprise me. There are so many moving parts in vLLM, that I do not expect any third-party component to successfully undergo comprehensive regression tests. That being said as Qwen 3.6 is still relevant model, it would be good to bring this to the attention of the KVarN authors.