---
cited_by:
  - '[[thoughts/papers/zhu2025hyperconnections#12-hc--hyper-connections-gated-residual-gr-variant|qwen4next §1.2]]'
  - '[[thoughts/inference/qwen3.8-serving#4-pr-reading-order-vllm-main-git-show-sha|qwen4next §4]]'
date: '2026-11-11'
kind: pr
local: []
merged: '2026-09-28'
number: 58957
source:
  - https://github.com/vllm-project/vllm/blob/d6cce94fd4221e6f5547e1c00ca5e0c92ab52d6c/vllm/models/qwen4_exp/nvidia/ops/cute_dsl/_hc_down_silu_mma.py
  - https://github.com/vllm-project/vllm/blob/d6cce94fd4221e6f5547e1c00ca5e0c92ab52d6c/vllm/models/qwen4_exp/nvidia/ops/cute_dsl/_hc_down_silu_fma.py
  - https://github.com/vllm-project/vllm/blob/d6cce94fd4221e6f5547e1c00ca5e0c92ab52d6c/vllm/models/qwen4_exp/nvidia/ops/cute_dsl/hc_down_silu.py
  - https://github.com/vllm-project/vllm/blob/d6cce94fd4221e6f5547e1c00ca5e0c92ab52d6c/vllm/models/qwen4_exp/nvidia/hyperconnection.py
  - https://github.com/vllm-project/vllm/blob/d6cce94fd4221e6f5547e1c00ca5e0c92ab52d6c/vllm/models/qwen4_exp/nvidia/model.py
  - https://github.com/vllm-project/vllm/blob/d6cce94fd4221e6f5547e1c00ca5e0c92ab52d6c/vllm/models/qwen4_exp/nvidia/ops/cute_dsl/__init__.py
state: merged
tags:
  - vllm
  - vllm-pr
  - new-architectures
title: '#58957 [Perf][Qwen4Exp] Fuse HC down projection and SiLU on NVIDIA'
topic: new-architectures
url: https://github.com/vllm-project/vllm/pull/58957
created: '2026-11-11'
modified: '2026-11-11'
published: '2026-11-11'
pageLayout: default
slug: hinterland/papers/maps/prs/58957
permalink: https://aarnphm.xyz/hinterland/papers/maps/prs/58957.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
Fuse Qwen4Exp’s BF16 HC down projection and SiLU for NVIDIA decode batches up to 48 tokens, with the existing unfused path for larger batches. This is separate from [#58706](https://github.com/aarnphm/aarnphm.github.io/issues/58706) (ROCm) and [#53909](https://github.com/aarnphm/aarnphm.github.io/issues/53909) (other HC operations). · full dump: [[thoughts/inference/vllm-prs/58957|#58957]]

