<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Horizon Labs on BriefTechNews.com</title><link>https://brief-tech-news.com/tags/horizon-labs/</link><description>Recent content in Horizon Labs 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>Wed, 18 Nov 2026 15:03:52 +0000</lastBuildDate><atom:link href="https://brief-tech-news.com/tags/horizon-labs/index.xml" rel="self" type="application/rss+xml"/><item><title>GPT-6.2 drops, Microsoft's $0.01/min model, and agents that build their own models</title><link>https://brief-tech-news.com/blogs/2026-11-18-ai/</link><pubDate>Wed, 18 Nov 2026 15:03:52 +0000</pubDate><guid>https://brief-tech-news.com/blogs/2026-11-18-ai/</guid><description>OpenAI shipped GPT-6.2 Astra and Sol, its frontier and everyday models, available now in ChatGPT Work, Codex, and the API. Microsoft answered with seven new MAI models at Ignite, including a cyber model that beats GPT-6 Astra&amp;rsquo;s quality at under 2% of the cost and a real-time voice model starting at a penny per audio minute, plus its Maia 200 accelerator. Replit launched Missions for agent swarms working within budgets, and two separate experiments showed LLMs writing neural network weights directly and an autonomous agent creating 38 adopted Hugging Face models. Runta&amp;rsquo;s new eval proves harness choice can swing cost by 5x, and Cerebras went production with disaggregated inference at 5x more tokens per megawatt.</description></item></channel></rss>