---
cited_by:
  - '[[thoughts/papers/kimiteam2026kimik3openfrontier#12-kda-kimi-delta-attention-layer-k3-variant|kimik3 §1.2]]'
  - '[[thoughts/inference/kimi-k3-state#33-spec-decode-with-recurrent-state|kimik3 §3.3]]'
  - '[[thoughts/inference/kimi-k3-prs#4-pr-reading-order|kimik3 §4]]'
date: '2026-11-11'
kind: pr
local:
  - file:///Users/aarnphm/workspace/vllm/vllm/utils/flashinfer.py
merged: '2026-10-01'
number: 54255
source:
  - https://github.com/vllm-project/vllm/blob/766b4ae54c491497306224ee4b316d9c2d57cd30/vllm/models/kimi_k3/nvidia/kda.py
  - https://github.com/vllm-project/vllm/blob/766b4ae54c491497306224ee4b316d9c2d57cd30/vllm/utils/flashinfer.py
state: merged
tags:
  - vllm
  - vllm-pr
  - new-architectures
title: '#54255 [Kimi-K3] Add FlashInfer speculative KDA backend'
topic: new-architectures
url: https://github.com/vllm-project/vllm/pull/54255
created: '2026-11-11'
modified: '2026-11-11'
published: '2026-11-11'
pageLayout: default
slug: hinterland/papers/maps/prs/54255
permalink: https://aarnphm.xyz/hinterland/papers/maps/prs/54255.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
Integrate FlashInfer’s packed fused KDA kernel into Kimi-K3 speculative decode. · full dump: [[thoughts/inference/vllm-prs/54255|#54255]]

