<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>xAI (Grok) on Bitsy Wiki</title>
    <link>https://wiki.bitsy.services/wiki/ai/models/xai/</link>
    <description>Recent content in xAI (Grok) on Bitsy Wiki</description>
    <generator>Hugo</generator>
    <language>en</language>
    <atom:link href="https://wiki.bitsy.services/wiki/ai/models/xai/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Grok 4.6</title>
      <link>https://wiki.bitsy.services/wiki/ai/models/xai/grok-4-6/</link>
      <pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate>
      <guid>https://wiki.bitsy.services/wiki/ai/models/xai/grok-4-6/</guid>
      <description>&lt;p&gt;Grok 4.6 is &lt;a href=&#34;https://wiki.bitsy.services/wiki/ai/models/xai&#34;&gt;xAI&lt;/a&gt;&amp;rsquo;s current generation, released 12 August 2026. Its announcement and model card both attribute it to &lt;strong&gt;SpaceXAI&lt;/strong&gt; rather than to xAI — the company was acquired by SpaceX in February 2026 and renamed in July, and this is the model where the new name appears on the vendor&amp;rsquo;s own pages.&lt;/p&gt;&#xA;&lt;p&gt;&lt;strong&gt;No architecture is published.&lt;/strong&gt; No parameter count, no layer count, no expert configuration. &lt;a href=&#34;https://wiki.bitsy.services/wiki/ai/models/xai/grok-1&#34;&gt;Grok-1&lt;/a&gt; in 2024 was the last generation xAI described, and everything since has been a product announcement rather than a technical one.&lt;/p&gt;</description>
    </item>
    <item>
      <title>Grok-1</title>
      <link>https://wiki.bitsy.services/wiki/ai/models/xai/grok-1/</link>
      <pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate>
      <guid>https://wiki.bitsy.services/wiki/ai/models/xai/grok-1/</guid>
      <description>&lt;p&gt;Grok-1 is the first model &lt;a href=&#34;https://wiki.bitsy.services/wiki/ai/models/xai&#34;&gt;xAI&lt;/a&gt; shipped, and the only one whose architecture it has ever published. Released 17 March 2024 under &lt;strong&gt;Apache 2.0&lt;/strong&gt; — code and weights both — it is a 314-billion-parameter &lt;a href=&#34;https://wiki.bitsy.services/wiki/ai/llm/mixture-of-experts&#34;&gt;mixture of experts&lt;/a&gt;, and at the time the largest open-weight model anyone had put out.&lt;/p&gt;&#xA;&lt;p&gt;It is on this wiki as a worked example rather than as a recommendation. Nothing xAI has released since discloses a parameter count, a layer count or a routing scheme, so Grok-1 is the last look inside the family. And because it is a full frontier-scale mixture of experts with every number published, it is a useful counterpart to &lt;a href=&#34;https://wiki.bitsy.services/wiki/ai/llm/gpt-2&#34;&gt;GPT-2&lt;/a&gt;, which is small and dense and runs on a laptop.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
