<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
  <title>MATRIX · openTPU</title>
  <link>https://derekwang2002.github.io/ai-intelligence/en/projects/opentpu/</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/opentpu.xml" rel="self" type="application/rss+xml"/>
  <item>
    <title>openTPU runs open models on a fully inspectable FPGA accelerator</title>
    <link>https://derekwang2002.github.io/ai-intelligence/en/events/ev-20260924-13/</link>
    <guid isPermaLink="false">ev-20260924-13-published</guid>
    <pubDate>Wed, 07 Oct 2026 22:04:39 GMT</pubDate>
    <category>New</category>
    <category>TRIAL</category>
    <description>&lt;p&gt;openTPU publishes an Apache 2.0 accelerator stack spanning SystemVerilog RTL, an instruction set, simulator, compiler, host tools and profiler. The project runs ten open models on a Kintex-7 PCIe FPGA and reports bit-for-bit agreement between the card and simulator. An October 7 update added Hugging Face golden-model validation and fixes from a whole-code assessment.&lt;/p&gt;&lt;p&gt;Recommendation: TRIAL&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://github.com/FeSens/openTPU&quot;&gt;FeSens/openTPU&lt;/a&gt;&lt;/p&gt;</description>
  </item>
</channel>
</rss>
