<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>local-models on Terry's Latent Space</title><link>https://taot.github.io/tags/local-models/</link><description>Recent content in local-models on Terry's Latent Space</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Mon, 19 Oct 2026 08:04:08 -0400</lastBuildDate><atom:link href="https://taot.github.io/tags/local-models/index.xml" rel="self" type="application/rss+xml"/><item><title>Strata Benchmark: Qwen3.8-Flash-Next Q2 on an RTX 5090</title><link>https://taot.github.io/posts/2026/10/strata-benchmark/</link><pubDate>Mon, 19 Oct 2026 07:54:25 -0400</pubDate><guid>https://taot.github.io/posts/2026/10/strata-benchmark/</guid><description>Quality and speed of Qwen3.8-Flash-Next (Q2_0) served by Strata on one RunPod RTX 5090, with thinking on and off.</description></item></channel></rss>