<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[MoE 模型是不是天然更适合企业部署]]></title><description><![CDATA[<p dir="auto">最近很多模型都在说 MoE。是不是 MoE 参数大但激活少，所以企业部署更划算？</p>
]]></description><link>https://localaihub.com/topic/147/moe-模型是不是天然更适合企业部署</link><generator>RSS for Node</generator><lastBuildDate>Wed, 03 Jun 2026 18:50:45 GMT</lastBuildDate><atom:link href="https://localaihub.com/topic/147.rss" rel="self" type="application/rss+xml"/><pubDate>Sun, 10 May 2026 06:49:00 GMT</pubDate><ttl>60</ttl><item><title><![CDATA[Reply to MoE 模型是不是天然更适合企业部署 on Mon, 11 May 2026 01:10:00 GMT]]></title><description><![CDATA[<p dir="auto">明白。MoE 是候选，不是答案。</p>
]]></description><link>https://localaihub.com/post/1491</link><guid isPermaLink="true">https://localaihub.com/post/1491</guid><dc:creator><![CDATA[小谢]]></dc:creator><pubDate>Mon, 11 May 2026 01:10:00 GMT</pubDate></item><item><title><![CDATA[Reply to MoE 模型是不是天然更适合企业部署 on Sun, 10 May 2026 23:29:00 GMT]]></title><description><![CDATA[<p dir="auto">还有许可证和生态。企业不是只买一次模型文件，后面要升级、回滚、排障。</p>
]]></description><link>https://localaihub.com/post/1490</link><guid isPermaLink="true">https://localaihub.com/post/1490</guid><dc:creator><![CDATA[ray_work]]></dc:creator><pubDate>Sun, 10 May 2026 23:29:00 GMT</pubDate></item><item><title><![CDATA[Reply to MoE 模型是不是天然更适合企业部署 on Sun, 10 May 2026 20:57:00 GMT]]></title><description><![CDATA[<p dir="auto">几乎所有模型选择问题最后都会回到评测集。</p>
]]></description><link>https://localaihub.com/post/1489</link><guid isPermaLink="true">https://localaihub.com/post/1489</guid><dc:creator><![CDATA[Grace]]></dc:creator><pubDate>Sun, 10 May 2026 20:57:00 GMT</pubDate></item><item><title><![CDATA[Reply to MoE 模型是不是天然更适合企业部署 on Sun, 10 May 2026 20:41:00 GMT]]></title><description><![CDATA[<p dir="auto">所以还是回到评测集。</p>
]]></description><link>https://localaihub.com/post/1488</link><guid isPermaLink="true">https://localaihub.com/post/1488</guid><dc:creator><![CDATA[小谢]]></dc:creator><pubDate>Sun, 10 May 2026 20:41:00 GMT</pubDate></item><item><title><![CDATA[Reply to MoE 模型是不是天然更适合企业部署 on Sun, 10 May 2026 19:52:00 GMT]]></title><description><![CDATA[<p dir="auto">如果要测，就拿同一批业务样例跑：延迟、吞吐、显存、错误率、格式稳定性。</p>
]]></description><link>https://localaihub.com/post/1487</link><guid isPermaLink="true">https://localaihub.com/post/1487</guid><dc:creator><![CDATA[小吴]]></dc:creator><pubDate>Sun, 10 May 2026 19:52:00 GMT</pubDate></item><item><title><![CDATA[Reply to MoE 模型是不是天然更适合企业部署 on Sun, 10 May 2026 18:04:00 GMT]]></title><description><![CDATA[<p dir="auto">对。一个模型在论文或榜单上强，不代表你们那台机器上能稳定跑。</p>
]]></description><link>https://localaihub.com/post/1486</link><guid isPermaLink="true">https://localaihub.com/post/1486</guid><dc:creator><![CDATA[阿航]]></dc:creator><pubDate>Sun, 10 May 2026 18:04:00 GMT</pubDate></item><item><title><![CDATA[Reply to MoE 模型是不是天然更适合企业部署 on Sun, 10 May 2026 15:15:00 GMT]]></title><description><![CDATA[<p dir="auto">我觉得 MoE 最大误区是“听起来省钱”。实际你还要能把它服务起来。</p>
]]></description><link>https://localaihub.com/post/1485</link><guid isPermaLink="true">https://localaihub.com/post/1485</guid><dc:creator><![CDATA[小满]]></dc:creator><pubDate>Sun, 10 May 2026 15:15:00 GMT</pubDate></item><item><title><![CDATA[Reply to MoE 模型是不是天然更适合企业部署 on Sun, 10 May 2026 14:02:00 GMT]]></title><description><![CDATA[<p dir="auto">风格稳定更多和训练、对齐、采样、提示词有关。不能只怪 MoE。</p>
]]></description><link>https://localaihub.com/post/1484</link><guid isPermaLink="true">https://localaihub.com/post/1484</guid><dc:creator><![CDATA[nora]]></dc:creator><pubDate>Sun, 10 May 2026 14:02:00 GMT</pubDate></item><item><title><![CDATA[Reply to MoE 模型是不是天然更适合企业部署 on Sun, 10 May 2026 13:31:00 GMT]]></title><description><![CDATA[<p dir="auto">MoE 会不会回答风格不稳定？</p>
]]></description><link>https://localaihub.com/post/1483</link><guid isPermaLink="true">https://localaihub.com/post/1483</guid><dc:creator><![CDATA[普通网友A]]></dc:creator><pubDate>Sun, 10 May 2026 13:31:00 GMT</pubDate></item><item><title><![CDATA[Reply to MoE 模型是不是天然更适合企业部署 on Sun, 10 May 2026 12:55:00 GMT]]></title><description><![CDATA[<p dir="auto">企业部署先看任务质量和稳定性，再看成本。模型架构不是采购理由。</p>
]]></description><link>https://localaihub.com/post/1482</link><guid isPermaLink="true">https://localaihub.com/post/1482</guid><dc:creator><![CDATA[melo]]></dc:creator><pubDate>Sun, 10 May 2026 12:55:00 GMT</pubDate></item><item><title><![CDATA[Reply to MoE 模型是不是天然更适合企业部署 on Sun, 10 May 2026 12:47:00 GMT]]></title><description><![CDATA[<p dir="auto">不一定。小并发、框架没优化、通信开销、显存放不下，都会让结果不理想。</p>
]]></description><link>https://localaihub.com/post/1481</link><guid isPermaLink="true">https://localaihub.com/post/1481</guid><dc:creator><![CDATA[林小北]]></dc:creator><pubDate>Sun, 10 May 2026 12:47:00 GMT</pubDate></item><item><title><![CDATA[Reply to MoE 模型是不是天然更适合企业部署 on Sun, 10 May 2026 11:43:00 GMT]]></title><description><![CDATA[<p dir="auto">那它不是一定比 dense 快？</p>
]]></description><link>https://localaihub.com/post/1480</link><guid isPermaLink="true">https://localaihub.com/post/1480</guid><dc:creator><![CDATA[小谢]]></dc:creator><pubDate>Sun, 10 May 2026 11:43:00 GMT</pubDate></item><item><title><![CDATA[Reply to MoE 模型是不是天然更适合企业部署 on Sun, 10 May 2026 11:35:00 GMT]]></title><description><![CDATA[<p dir="auto">“总参数大、激活参数少”是真的方向，但落到部署要看具体模型和后端支持。</p>
]]></description><link>https://localaihub.com/post/1479</link><guid isPermaLink="true">https://localaihub.com/post/1479</guid><dc:creator><![CDATA[陈一]]></dc:creator><pubDate>Sun, 10 May 2026 11:35:00 GMT</pubDate></item><item><title><![CDATA[Reply to MoE 模型是不是天然更适合企业部署 on Sun, 10 May 2026 08:52:00 GMT]]></title><description><![CDATA[<p dir="auto">别直接等号。MoE 的推理成本和显存、路由、并发、框架支持都有关系。</p>
]]></description><link>https://localaihub.com/post/1478</link><guid isPermaLink="true">https://localaihub.com/post/1478</guid><dc:creator><![CDATA[阿白]]></dc:creator><pubDate>Sun, 10 May 2026 08:52:00 GMT</pubDate></item></channel></rss>