---
cited_by:
  - '[[thoughts/inference/replayssm|replayssm]]'
date: '2026-11-11'
kind: rfc
local: []
number: 47572
source: []
state: open
tags:
  - vllm
  - vllm-pr
  - new-architectures
title: '#47572 [RFC]: ReplaySSM: cache SSM inputs instead of state for faster standard and speculative decode (Mamba2 + GDN)'
topic: new-architectures
url: https://github.com/vllm-project/vllm/issues/47572
created: '2026-11-11'
modified: '2026-11-11'
published: '2026-11-11'
pageLayout: default
slug: hinterland/papers/maps/prs/47572
permalink: https://aarnphm.xyz/hinterland/papers/maps/prs/47572.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
ReplaySSM: cache SSM inputs instead of state for faster standard and speculative decode (Mamba2 + GDN) · full dump: [[thoughts/inference/vllm-prs/47572|#47572]]

