<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[Read this first: what the benchmark actually measures]]></title><description><![CDATA[<p dir="auto">The benchmark runs your model on <strong>your</strong> hardware and reports what happened. A few principles it sticks to:</p>
<ul>
<li><strong>Measured means measured.</strong> Every number is labelled — <code>Measured</code>, <code>Estimated</code>, <code>Mixed</code>, <code>Partial</code>. If a run didn't finish, you get <code>Incomplete</code>, not a filled-in guess.</li>
<li><strong>Nothing is invented.</strong> There are no synthetic values anywhere. A missing number shows as <code>—</code>.</li>
<li><strong>Derived numbers inherit the weakest input.</strong> Measured speed divided by an estimated memory figure is labelled <code>Mixed</code>, never <code>Measured</code>.</li>
</ul>
<p dir="auto"><strong>Free readiness check vs. full benchmark:</strong> the readiness check in setup is free and always available — it tells you tokens/sec for your model on this machine. The full suite (Lizard vs Ollama vs llama.cpp across Q1–Q9) is the credit-based one.</p>
<p dir="auto">Full explanation: <a href="https://lizard-llm.qendryx.com/benchmarks.html" rel="nofollow ugc">https://lizard-llm.qendryx.com/benchmarks.html</a></p>
<p dir="auto">When posting results, your CPU/GPU and model + quantization make them useful to everyone else.</p>
]]></description><link>https://community.lizard-llm.qendryx.com/topic/11/read-this-first-what-the-benchmark-actually-measures</link><generator>RSS for Node</generator><lastBuildDate>Mon, 24 Aug 2026 03:19:39 GMT</lastBuildDate><atom:link href="https://community.lizard-llm.qendryx.com/topic/11.rss" rel="self" type="application/rss+xml"/><pubDate>Sat, 11 Jul 2026 10:50:00 GMT</pubDate><ttl>60</ttl></channel></rss>