<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
  <title>MATRIX · Muse Spark Open</title>
  <link>https://derekwang2002.github.io/ai-intelligence/en/projects/muse-spark/</link>
  <description>Material changes in AI models, agents, engineering and infrastructure, with primary sources and recommendations.</description>
  <language>en</language>
  <lastBuildDate>Mon, 02 Nov 2026 16:56:00 GMT</lastBuildDate>
  <atom:link href="https://derekwang2002.github.io/ai-intelligence/en/feeds/projects/muse-spark.xml" rel="self" type="application/rss+xml"/>
  <item>
    <title>Muse Spark moves from an open-weight preview to downloadable and hosted access</title>
    <link>https://derekwang2002.github.io/ai-intelligence/en/events/ev-20260902-08/</link>
    <guid isPermaLink="false">ev-20260902-08-muse-spark-open</guid>
    <pubDate>Thu, 15 Oct 2026 16:00:25 GMT</pubDate>
    <category>Update</category>
    <category>TRIAL</category>
    <description>&lt;p&gt;Muse Spark moves from an open-weight preview to downloadable and hosted access&lt;/p&gt;&lt;p&gt;Original story: Meta releases Muse Spark Open for local and hosted inference&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Why read: &lt;/strong&gt;Muse Spark moves from a hosted preview to open weights and multiple runtimes, while its 816B size, gating and custom license still materially constrain adoption.&lt;/p&gt;&lt;p&gt;Recommendation: TRIAL&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://huggingface.co/blog/meta-muse-spark-open&quot;&gt;Muse Spark Open&lt;/a&gt;&lt;/p&gt;</description>
  </item>
  <item>
    <title>Meta releases Muse Spark Open for local and hosted inference</title>
    <link>https://derekwang2002.github.io/ai-intelligence/en/events/ev-20260902-08/</link>
    <guid isPermaLink="false">ev-20260902-08-published</guid>
    <pubDate>Sat, 05 Sep 2026 00:00:21 GMT</pubDate>
    <category>New</category>
    <category>TRIAL</category>
    <description>&lt;p&gt;Meta released the gated Muse Spark Open weights, fulfilling the earlier open-weight preview. The 816B-parameter mixture-of-experts model activates 42B parameters per token, accepts text, images, audio and video, and has a one-million-token context window. Reference paths include Hugging Face endpoints, SGLang, vLLM, llama.cpp and oMLX. The custom license requires attribution and a separate agreement for model-as-a-service businesses above $20 million in annual gross revenue.&lt;/p&gt;&lt;p&gt;Recommendation: TRIAL&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://research.meta.ai/blog/introducing-muse-spark-1-3&quot;&gt;research.meta.ai · introducing muse spark 1 3&lt;/a&gt;&lt;/p&gt;</description>
  </item>
</channel>
</rss>
