---
cited_by:
  - '[[thoughts/inference/kimi-k3#5-sharp-edges-each-fix-shows-an-invariant|kimik3 §5]]'
date: '2026-11-11'
kind: pr
local: []
merged: '2026-08-22'
number: 53327
source:
  - https://github.com/vllm-project/vllm/blob/e9d1398d9edfd90fcc1cf783805240e3effec013/vllm/models/kimi_k3/nvidia/latent_moe_runner.py
state: merged
tags:
  - vllm
  - vllm-pr
  - new-architectures
title: '#53327 [Bugfix][Kimi K3] Enable deferred MoE finalization before weight loading'
topic: new-architectures
url: https://github.com/vllm-project/vllm/pull/53327
created: '2026-11-11'
modified: '2026-11-11'
published: '2026-11-11'
pageLayout: default
slug: hinterland/papers/maps/prs/53327
permalink: https://aarnphm.xyz/hinterland/papers/maps/prs/53327.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
Enable K3 deferred MoE finalization from the selected monolithic expert class during runner construction. · full dump: [[thoughts/inference/vllm-prs/53327|#53327]]

