---
type: "post"
id: "engineers-at-humans-accelerated-mixture-of-experts-decode-840662"
slug: "engineers-at-humans-accelerated-mixture-of-experts-decode-840662"
title: "Engineers at humans& accelerated Mixture-of-Experts decode on NVIDIA Blackwell GPUs by keeping weights in 4-bit NVFP4 while retaining activations in 16-bit BF16."
description: "humans& reported a Mixture-of-Experts decode kernel for NVIDIA Blackwell GPUs that keeps static weights in 4-bit NVFP4 while retaining activations in BF16."
retrieval_nugget: "humans& reported a Mixture-of-Experts decode kernel for NVIDIA Blackwell GPUs that keeps static weights in 4-bit NVFP4 while retaining activations in BF16."
published_at: "2026-10-25"
updated_at: "2026-10-25"
record_date: "2026-10-25"
date_kind: "published_at"
topics: ["inference","compute"]
entities: ["humans&","NVIDIA Blackwell","NVFP4","BF16","FlashInfer","MegaMoE","GLM-5","NVIDIA B300"]
related_patterns: []
novelty_label: "new_fact"
what_changed: "humans& reports a distinct Blackwell MoE decoding technique: NVFP4 weights with BF16 activations improve both speed and accuracy."
prior_refs: ["memory:feb9e401-6837-8ff3-adf1-2e4ddc6c644b"]
source_url: "https://humansand.ai/blog/faster-than-light-nvfp4-decode"
source_urls: ["https://humansand.ai/blog/faster-than-light-nvfp4-decode"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:engineers-at-humans-accelerated-mixture-of-experts-decode-840662"
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/engineers-at-humans-accelerated-mixture-of-experts-decode-840662/","markdown":"https://newruntime.com/posts/engineers-at-humans-accelerated-mixture-of-experts-decode-840662.md","json":"https://newruntime.com/posts/engineers-at-humans-accelerated-mixture-of-experts-decode-840662.json"}
---

# Engineers at humans& accelerated Mixture-of-Experts decode on NVIDIA Blackwell GPUs by keeping weights in 4-bit NVFP4 while retaining activations in 16-bit BF16.

## Retrieval answer

humans& reported a Mixture-of-Experts decode kernel for NVIDIA Blackwell GPUs that keeps static weights in 4-bit NVFP4 while retaining activations in BF16.

humans& reported a Mixture-of-Experts decode kernel for NVIDIA Blackwell GPUs that keeps static weights in 4-bit NVFP4 while retaining activations in BF16. The company calls the approach W4A16, contrasting it with conventional W4A4 paths that quantize activations during decode.

According to the October 23, 2026 post, the kernel removes activation quantization from the critical path and overlaps weight dequantization with computation and data movement. It uses Blackwell features including native NVFP4 conversion instructions, tensor memory, asynchronous execution, and Tensor Memory Accelerator and matrix-multiply-accumulate instruction handling. humans& says its implementation outperformed FlashInfer's optimized W4A4 MegaMoE path by up to 25% in a GLM-5-shaped layer test across eight B300 GPUs and 600 samples per configuration.

The source presents a company benchmark, not an independent evaluation. It does not establish end-to-end serving gains for a named production model, performance on GPU generations other than Blackwell, cost changes, or general accuracy results across workloads.

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: [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)