<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0"><channel><title>A New Open-Weight Challenger to Anthropic, Reflection, Emerges — Live Feed</title><link>https://www.live-feeds.com/feed/a-new-open-weight-challenger-to-anthropic-reflection-emerges</link><atom:link xmlns:atom="http://www.w3.org/2005/Atom" href="https://www.live-feeds.com/feed/a-new-open-weight-challenger-to-anthropic-reflection-emerges/rss.xml" rel="self" type="application/rss+xml"/><description>Continuously updated, source-cited coverage.</description>
<item><title>Nvidia-Backed Reflection AI Unveils Beam Open-Weight Model</title><link>https://www.live-feeds.com/feed/a-new-open-weight-challenger-to-anthropic-reflection-emerges</link><guid isPermaLink="false">https://www.live-feeds.com/feed/a-new-open-weight-challenger-to-anthropic-reflection-emerges#u112181</guid><pubDate>Fri, 09 Oct 2026 13:26:34 +0000</pubDate><description>Reflection AI, an American start-up founded by Google DeepMind alumni and backed by Nvidia, has introduced its first open-weight artificial intelligence model named Beam. The 501-billion-parameter sparse Mixture-of-Experts model features 23 billion active parameters and is designed to handle coding, reasoning, and agentic workloads. Beam was trained on 23.8 trillion tokens utilizing curated web and proprietary licensed datasets, while its reinforcement learning phase used over 100 million rollouts across 10,500 Nvidia GB300 graphics processing units during a four-week training run.Why it matte</description></item>
</channel></rss>