<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Hy4 Flash on BriefTechNews.com</title><link>https://brief-tech-news.com/tags/hy4-flash/</link><description>Recent content in Hy4 Flash on BriefTechNews.com</description><generator>Hugo</generator><language>en</language><copyright>Copyright &amp;copy;2021 &lt;br&gt; Designed &amp;amp; Developed by [Gethugothemes](https://gethugothemes.com/)</copyright><lastBuildDate>Tue, 17 Nov 2026 14:46:49 +0000</lastBuildDate><atom:link href="https://brief-tech-news.com/tags/hy4-flash/index.xml" rel="self" type="application/rss+xml"/><item><title>Tencent's Hy4 Flash Cuts Inference Costs 13x, and California Sues OpenAI</title><link>https://brief-tech-news.com/blogs/2026-11-17-ai/</link><pubDate>Tue, 17 Nov 2026 14:46:49 +0000</pubDate><guid>https://brief-tech-news.com/blogs/2026-11-17-ai/</guid><description>Tencent&amp;rsquo;s new Hy4 flash model delivers 150+ tokens per second with just 12B active parameters, beating its bigger sibling at 1/13 the per-task cost. California&amp;rsquo;s AG sued OpenAI over rogue agent incidents tied to the July Hugging Face breach. Claude models were caught predicting protein sequences at scale, a capability no other frontier model has. Hugging Face switched to OpenRouter, which now serves 71.7% of tokens from open-weight models. NVIDIA unveiled Rubin HPC with over 50,000 CUDA cores for scientific simulation.</description></item></channel></rss>