---
cited_by:
  - '[[thoughts/papers/qiu2026designqwen38nextarchitectureevaluation#11-qsa--qwen-sparse-attention|qwen4next §1.1]]'
  - '[[thoughts/inference/qwen3.8-serving#4-pr-reading-order-vllm-main-git-show-sha|qwen4next §4]]'
date: '2026-11-11'
kind: pr
local: []
merged: '2026-09-16'
number: 55557
source:
  - https://github.com/vllm-project/vllm/blob/dff1bde84dd6e34a49c150116d0f212507280910/vllm/models/qwen4_exp/nvidia/ops/qsa.py
  - https://github.com/vllm-project/vllm/blob/dff1bde84dd6e34a49c150116d0f212507280910/vllm/models/qwen4_exp/nvidia/qsa.py
state: merged
tags:
  - vllm
  - vllm-pr
  - new-architectures
title: '#55557 [Model] Qwen4Exp: fp8_e4m3 main KV cache on the QSA path'
topic: new-architectures
url: https://github.com/vllm-project/vllm/pull/55557
created: '2026-11-11'
modified: '2026-11-11'
published: '2026-11-11'
pageLayout: default
slug: hinterland/papers/maps/prs/55557
permalink: https://aarnphm.xyz/hinterland/papers/maps/prs/55557.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
Let Qwen3.8-Flash-Next store the main K/V cache of its QSA (sparse attention) layers in · full dump: [[thoughts/inference/vllm-prs/55557|#55557]]

