<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>latentwire: Models</title><link>https://latentwire.enphorm.com/</link><description>Models - New model releases, capabilities, and benchmark claims.</description><language>en</language><lastBuildDate>Mon, 07 Sep 2026 01:38:33 +0000</lastBuildDate><atom:link href="https://latentwire.enphorm.com/models/rss.xml" rel="self" type="application/rss+xml"/><item><title>OpenAI launches GPT-6 Astra, claims AGI-era computer use and coding gains</title><link>https://latentwire.enphorm.com/stories/openai-launches-gpt-6-astra-claims-agi-era-computer-use-and-559cf3/</link><guid isPermaLink="true">https://latentwire.enphorm.com/stories/openai-launches-gpt-6-astra-claims-agi-era-computer-use-and-559cf3/</guid><pubDate>Sun, 06 Sep 2026 12:01:51 +0000</pubDate><category>Models</category><description>GPT-6 Astra marks OpenAI's most aggressive push yet into agentic computer use and autonomous security work, while its opaque reasoning and benchmark caveats raise open questions about how much of the claimed capability is real versus harness-dependent.</description></item><item><title>Qwen3.8-Flash-Next: 125B MoE previews Qwen4 architecture, runs locally</title><link>https://latentwire.enphorm.com/stories/qwen3-8-flash-next-125b-moe-previews-qwen4-architecture-runs-461c76/</link><guid isPermaLink="true">https://latentwire.enphorm.com/stories/qwen3-8-flash-next-125b-moe-previews-qwen4-architecture-runs-461c76/</guid><pubDate>Sun, 06 Sep 2026 12:02:43 +0000</pubDate><category>Models</category><description>A 125B-parameter MoE with only 6B active makes near-frontier multimodal capability practical on local hardware while previewing Qwen4's architecture.</description></item></channel></rss>