<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
  <title>MATRIX · Recursive Flint</title>
  <link>https://derekwang2002.github.io/ai-intelligence/en/projects/recursive-flint/</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/recursive-flint.xml" rel="self" type="application/rss+xml"/>
  <item>
    <title>Recursive releases Flint as an open-weight sparse coding model</title>
    <link>https://derekwang2002.github.io/ai-intelligence/en/events/ev-20261021-12/</link>
    <guid isPermaLink="false">ev-20261021-12-published</guid>
    <pubDate>Thu, 22 Oct 2026 16:37:30 GMT</pubDate>
    <category>New</category>
    <category>TRIAL</category>
    <description>&lt;p&gt;Recursive released Flint, a 39.8B-parameter mixture-of-experts coding model with about 2.5B active parameters per token and a 262K context window. BF16 weights are Apache-2.0, with SGLang and vLLM instructions; Ollama 0.40.8 adds a roughly 23 GB package for macOS. Other Ollama GPU support is promised later, and benchmarks remain publisher-run.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Why read: &lt;/strong&gt;Apache-2.0 weights and Ollama/vLLM paths make a controlled local coding-model comparison possible now.&lt;/p&gt;&lt;p&gt;Recommendation: TRIAL&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://github.com/ollama/ollama/releases/tag/v0.40.8&quot;&gt;Ollama 0.40.8 release&lt;/a&gt; · &lt;a href=&quot;https://huggingface.co/RecursiveSI/flint&quot;&gt;Recursive Flint model card&lt;/a&gt; · &lt;a href=&quot;https://ollama.com/library/flint&quot;&gt;Ollama Flint library page&lt;/a&gt;&lt;/p&gt;</description>
  </item>
</channel>
</rss>
