---
type: "post"
id: "disaggregated-deployments-can-assign-precision-by-serving-phase-01b3fb"
slug: "disaggregated-deployments-can-assign-precision-by-serving-phase-01b3fb"
title: "Disaggregated deployments can assign precision by serving phase because both paths consume the same prepared NVFP4 weight layout."
description: "FlashInfer pull request #5319 proposes an optimization for the unfused CuTe DSL MoE finalize path."
retrieval_nugget: "FlashInfer pull request #5319 proposes an optimization for the unfused CuTe DSL MoE finalize path."
published_at: "2026-10-25"
updated_at: "2026-10-25"
record_date: "2026-10-25"
date_kind: "published_at"
topics: ["inference","ai-infrastructure","compute"]
entities: ["FlashInfer","CuTe DSL","NVIDIA B300"]
related_patterns: []
novelty_label: "new_angle"
what_changed: "The FlashInfer change offers a shared prepared NVFP4 weight layout for phase-specific serving precision, but describes kernel implementation rather than the prior papers’ quantization approach."
prior_refs: ["memory:20682048-be1f-868d-a70b-60a37aff5b6c","memory:dbf688b1-4b4f-82e3-a6a0-4cd6b08e021c"]
source_url: "https://github.com/flashinfer-ai/flashinfer/pull/5319"
source_urls: ["https://github.com/flashinfer-ai/flashinfer/pull/5319"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:disaggregated-deployments-can-assign-precision-by-serving-phase-01b3fb"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/disaggregated-deployments-can-assign-precision-by-serving-phase-01b3fb/","markdown":"https://newruntime.com/posts/disaggregated-deployments-can-assign-precision-by-serving-phase-01b3fb.md","json":"https://newruntime.com/posts/disaggregated-deployments-can-assign-precision-by-serving-phase-01b3fb.json"}
---

# Disaggregated deployments can assign precision by serving phase because both paths consume the same prepared NVFP4 weight layout.

## Retrieval answer

FlashInfer pull request #5319 proposes an optimization for the unfused CuTe DSL MoE finalize path.

FlashInfer pull request #5319 proposes an optimization for the unfused CuTe DSL MoE finalize path. It covers W4A4 NVFP4 activations with per-tensor or per-token scaling and W4A16 with BF16 activations.

The change replaces a standalone native finalizer with a shared CuTe kernel, adjusts dependent-launch hints, and wires the `--no-fused-finalize` benchmark flag through to CUTLASS. The stated scope is six CuTe DSL and benchmark-wiring files, with no C++ changes; public MoE defaults and fused-finalize arithmetic are unchanged.

The PR records a historical single-sweep benchmark on one NVIDIA B300 GPU: all 15 tested token counts improved for each of W4A4 per-tensor, W4A4 per-token, and W4A16. After a September 30 rebase, the author reports validation checks but says GPU benchmarks were not rerun. The source does not establish that the PR has merged, shipped in a FlashInfer release, or that its historical performance results apply to current upstream code.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Inference](https://newruntime.com/topics/inference.md): Explore the inference topic hub.
- topic hub: [Ai Infrastructure](https://newruntime.com/topics/ai-infrastructure.md): Explore the ai-infrastructure topic hub.
- topic hub: [Compute](https://newruntime.com/topics/compute.md): Explore the compute topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)