---
cited_by:
  - '[[thoughts/inference/replayssm-serving#5-pr-reading-order|replayssm §5]]'
date: '2026-11-11'
kind: pr
local: []
merged: '2026-08-11'
number: 51819
source:
  - https://github.com/vllm-project/vllm/blob/0fb9897df12963e9bd0d58547fc997e0702d95ec/vllm/model_executor/layers/fused_moe/experts/flashinfer_b12x_moe.py
state: merged
tags:
  - vllm
  - vllm-pr
  - new-architectures
title: '#51819 [Bugfix][MoE] Support GELU tanh in FlashInfer B12x MoE'
topic: new-architectures
url: https://github.com/vllm-project/vllm/pull/51819
created: '2026-11-11'
modified: '2026-11-11'
published: '2026-11-11'
pageLayout: default
slug: hinterland/papers/maps/prs/51819
permalink: https://aarnphm.xyz/hinterland/papers/maps/prs/51819.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
Adds `MoEActivation.GELU_TANH` support to the FlashInfer B12x MoE backend so Gemma 4 NVFP4 can use `--moe-backend flashinfer_b12x`. · full dump: [[thoughts/inference/vllm-prs/51819|#51819]]

