<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>PostNAS on AI内参</title>
    <link>https://www.neican.ai/tags/postnas/</link>
    <description>Recent content in PostNAS on AI内参</description>
    <generator>Hugo</generator>
    <language>zh-cn</language>
    <lastBuildDate>Wed, 27 Aug 2025 10:10:04 +0800</lastBuildDate>
    <atom:link href="https://www.neican.ai/tags/postnas/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>英伟达“小钢炮”炸场：推理狂飙53倍，还把Mamba 2甩在了身后，这波操作太“顶”了！</title>
      <link>https://www.neican.ai/insights/53mamba-2-20250827101004698-1/</link>
      <pubDate>Wed, 27 Aug 2025 10:10:04 +0800</pubDate>
      <guid>https://www.neican.ai/insights/53mamba-2-20250827101004698-1/</guid>
      <description>英伟达最近发布了“小而精”的Jet-Nemotron系列小模型，不仅在H100 GPU上实现了最高53倍的推理吞吐量提升，还在多项基准测试中超越了Qwen3、Llama3.2等主流模型，甚至在架构上超越了Mamba 2。其核心黑科技是PostNAS（后神经网络架构搜索）和JetBlock（新型线性注意力模块），让模型在保持高精度的同时，运行效率飙升，预示着英伟达在小模型赛道上祭出了“王炸”！</description>
    </item>
  </channel>
</rss>
