<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
  <channel>
    <title>Qwen on blog.byMAR.CO</title>
    <link>https://blog.bymar.co/tags/qwen/</link>
    <description>Recent content in Qwen on blog.byMAR.CO</description>
    <generator>Hugo -- 0.161.1</generator>
    <language>en-us</language>
    <lastBuildDate>Thu, 07 May 2026 02:10:00 -0400</lastBuildDate>
    <atom:link href="https://blog.bymar.co/tags/qwen/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Speculative Decoding Is Finally Useful for Local LLMs</title>
      <link>https://blog.bymar.co/posts/speculative-decoding-local-llms-2026/</link>
      <pubDate>Thu, 07 May 2026 02:10:00 -0400</pubDate>
      <guid>https://blog.bymar.co/posts/speculative-decoding-local-llms-2026/</guid>
      <description>A practical field guide to speculative decoding for local LLMs, covering ngram speculation, MTP heads, TurboQuant, Qwen3.6, llama.cpp, vLLM, and the current LocalLLaMA field reports.</description>
    </item>
  </channel>
</rss>
