<?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 Pavel Nasovich's Blog</title><link>https://forcewake.me/tags/local-models/</link><description>Recent content in Local-Models on Pavel Nasovich's Blog</description><generator>Hugo -- 0.157.0</generator><language>en-us</language><copyright>Copyright 2026</copyright><lastBuildDate>Sun, 05 Apr 2026 23:39:32 +0200</lastBuildDate><atom:link href="https://forcewake.me/tags/local-models/index.xml" rel="self" type="application/rss+xml"/><item><title>I Ran pi on Gemma 4 26B A4B via llama.cpp. Here Is What Broke First</title><link>https://forcewake.me/pi-gemma4-llamacpp-local-coding-agent/</link><pubDate>Sat, 04 Apr 2026 19:12:20 +0000</pubDate><guid>https://forcewake.me/pi-gemma4-llamacpp-local-coding-agent/</guid><description>A first-person technical teardown of running pi against a local llama.cpp server on Unsloth&amp;#39;s Gemma 4 26B A4B GGUF, including what worked, what failed, and why 4-bit precision becomes the real bottleneck.</description></item></channel></rss>