<?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[选模型最后卡在“大家都说不错”，怎么做决策？]]></title><description><![CDATA[<p dir="auto">我们试了 Qwen、DeepSeek、Llama、Gemma、GLM、Kimi、GPT、Claude，感觉都不错。最后怎么拍板？</p>
]]></description><link>https://localaihub.com/topic/102/选模型最后卡在-大家都说不错-怎么做决策</link><generator>RSS for Node</generator><lastBuildDate>Wed, 03 Jun 2026 17:50:58 GMT</lastBuildDate><atom:link href="https://localaihub.com/topic/102.rss" rel="self" type="application/rss+xml"/><pubDate>Wed, 06 May 2026 14:44:00 GMT</pubDate><ttl>60</ttl><item><title><![CDATA[Reply to 选模型最后卡在“大家都说不错”，怎么做决策？ on Thu, 07 May 2026 15:29:00 GMT]]></title><description><![CDATA[<p dir="auto">这就是生产决策。模型卡看边界，评测集看你自己的业务。</p>
]]></description><link>https://localaihub.com/post/814</link><guid isPermaLink="true">https://localaihub.com/post/814</guid><dc:creator><![CDATA[陈一]]></dc:creator><pubDate>Thu, 07 May 2026 15:29:00 GMT</pubDate></item><item><title><![CDATA[Reply to 选模型最后卡在“大家都说不错”，怎么做决策？ on Thu, 07 May 2026 13:10:00 GMT]]></title><description><![CDATA[<p dir="auto">我们按场景建决策矩阵，不再争一个“最好模型”。第一版中文客服主路径先选成本和中文稳定性最优的，复杂问题再升级。</p>
]]></description><link>https://localaihub.com/post/813</link><guid isPermaLink="true">https://localaihub.com/post/813</guid><dc:creator><![CDATA[麦田]]></dc:creator><pubDate>Thu, 07 May 2026 13:10:00 GMT</pubDate></item><item><title><![CDATA[Reply to 选模型最后卡在“大家都说不错”，怎么做决策？ on Thu, 07 May 2026 12:21:00 GMT]]></title><description><![CDATA[<p dir="auto">很可能。但先选主路径，再加清晰兜底。多模型不是堆模型，是可解释路由。</p>
]]></description><link>https://localaihub.com/post/812</link><guid isPermaLink="true">https://localaihub.com/post/812</guid><dc:creator><![CDATA[林小北]]></dc:creator><pubDate>Thu, 07 May 2026 12:21:00 GMT</pubDate></item><item><title><![CDATA[Reply to 选模型最后卡在“大家都说不错”，怎么做决策？ on Thu, 07 May 2026 10:05:00 GMT]]></title><description><![CDATA[<p dir="auto">会不会最后变成多模型？</p>
]]></description><link>https://localaihub.com/post/811</link><guid isPermaLink="true">https://localaihub.com/post/811</guid><dc:creator><![CDATA[小周]]></dc:creator><pubDate>Thu, 07 May 2026 10:05:00 GMT</pubDate></item><item><title><![CDATA[Reply to 选模型最后卡在“大家都说不错”，怎么做决策？ on Thu, 07 May 2026 07:36:00 GMT]]></title><description><![CDATA[<p dir="auto">供应商和部署风险也要算。API 限额、区域、日志、模型版本变化、本地资源，都是真成本。</p>
]]></description><link>https://localaihub.com/post/810</link><guid isPermaLink="true">https://localaihub.com/post/810</guid><dc:creator><![CDATA[rootless]]></dc:creator><pubDate>Thu, 07 May 2026 07:36:00 GMT</pubDate></item><item><title><![CDATA[Reply to 选模型最后卡在“大家都说不错”，怎么做决策？ on Thu, 07 May 2026 05:47:00 GMT]]></title><description><![CDATA[<p dir="auto">还要让真实使用者参与。工程师觉得强，客服同事可能觉得语气奇怪。</p>
]]></description><link>https://localaihub.com/post/809</link><guid isPermaLink="true">https://localaihub.com/post/809</guid><dc:creator><![CDATA[葡萄冰]]></dc:creator><pubDate>Thu, 07 May 2026 05:47:00 GMT</pubDate></item><item><title><![CDATA[Reply to 选模型最后卡在“大家都说不错”，怎么做决策？ on Thu, 07 May 2026 03:50:00 GMT]]></title><description><![CDATA[<p dir="auto">按业务风险。客服退款，严重错误率权重大；内部写作，成本和风格更重要；代码 agent，测试通过和 diff 质量更重要。</p>
]]></description><link>https://localaihub.com/post/808</link><guid isPermaLink="true">https://localaihub.com/post/808</guid><dc:creator><![CDATA[阿航]]></dc:creator><pubDate>Thu, 07 May 2026 03:50:00 GMT</pubDate></item><item><title><![CDATA[Reply to 选模型最后卡在“大家都说不错”，怎么做决策？ on Thu, 07 May 2026 02:52:00 GMT]]></title><description><![CDATA[<p dir="auto">权重怎么定？</p>
]]></description><link>https://localaihub.com/post/807</link><guid isPermaLink="true">https://localaihub.com/post/807</guid><dc:creator><![CDATA[普通网友A]]></dc:creator><pubDate>Thu, 07 May 2026 02:52:00 GMT</pubDate></item><item><title><![CDATA[Reply to 选模型最后卡在“大家都说不错”，怎么做决策？ on Thu, 07 May 2026 00:29:00 GMT]]></title><description><![CDATA[<p dir="auto">决策矩阵别太花。准确率、严重错误率、成本、延迟、可部署性、合规、维护成本。六七项够了。</p>
]]></description><link>https://localaihub.com/post/806</link><guid isPermaLink="true">https://localaihub.com/post/806</guid><dc:creator><![CDATA[zeroOne]]></dc:creator><pubDate>Thu, 07 May 2026 00:29:00 GMT</pubDate></item><item><title><![CDATA[Reply to 选模型最后卡在“大家都说不错”，怎么做决策？ on Wed, 06 May 2026 22:01:00 GMT]]></title><description><![CDATA[<p dir="auto">GLM 可以在中文正式材料、企业文档里测一测，别只拿客服口语评判它。</p>
]]></description><link>https://localaihub.com/post/805</link><guid isPermaLink="true">https://localaihub.com/post/805</guid><dc:creator><![CDATA[leaf_1997]]></dc:creator><pubDate>Wed, 06 May 2026 22:01:00 GMT</pubDate></item><item><title><![CDATA[Reply to 选模型最后卡在“大家都说不错”，怎么做决策？ on Wed, 06 May 2026 21:38:00 GMT]]></title><description><![CDATA[<p dir="auto">Kimi/Claude/GPT 这类强长上下文或通用模型适合复杂材料和兜底，但成本、延迟、权限要控。</p>
]]></description><link>https://localaihub.com/post/804</link><guid isPermaLink="true">https://localaihub.com/post/804</guid><dc:creator><![CDATA[nora]]></dc:creator><pubDate>Wed, 06 May 2026 21:38:00 GMT</pubDate></item><item><title><![CDATA[Reply to 选模型最后卡在“大家都说不错”，怎么做决策？ on Wed, 06 May 2026 20:41:00 GMT]]></title><description><![CDATA[<p dir="auto">DeepSeek 适合复杂推理和代码分析候选，但别所有短任务都上推理。</p>
]]></description><link>https://localaihub.com/post/803</link><guid isPermaLink="true">https://localaihub.com/post/803</guid><dc:creator><![CDATA[melo]]></dc:creator><pubDate>Wed, 06 May 2026 20:41:00 GMT</pubDate></item><item><title><![CDATA[Reply to 选模型最后卡在“大家都说不错”，怎么做决策？ on Wed, 06 May 2026 20:14:00 GMT]]></title><description><![CDATA[<p dir="auto">Qwen 可做中文和本地优先候选，Llama 看英文生态和本地经验，Gemma 看轻量任务。</p>
]]></description><link>https://localaihub.com/post/802</link><guid isPermaLink="true">https://localaihub.com/post/802</guid><dc:creator><![CDATA[小高]]></dc:creator><pubDate>Wed, 06 May 2026 20:14:00 GMT</pubDate></item><item><title><![CDATA[Reply to 选模型最后卡在“大家都说不错”，怎么做决策？ on Wed, 06 May 2026 18:38:00 GMT]]></title><description><![CDATA[<p dir="auto">先按场景分：中文客服、代码助手、长文阅读、轻量分类、内部分析。不要选一个总冠军。</p>
]]></description><link>https://localaihub.com/post/801</link><guid isPermaLink="true">https://localaihub.com/post/801</guid><dc:creator><![CDATA[陈一]]></dc:creator><pubDate>Wed, 06 May 2026 18:38:00 GMT</pubDate></item><item><title><![CDATA[Reply to 选模型最后卡在“大家都说不错”，怎么做决策？ on Wed, 06 May 2026 16:54:00 GMT]]></title><description><![CDATA[<p dir="auto">“都不错”说明样本太简单。加失败样本、高风险样本、真实噪声样本。</p>
]]></description><link>https://localaihub.com/post/800</link><guid isPermaLink="true">https://localaihub.com/post/800</guid><dc:creator><![CDATA[林小北]]></dc:creator><pubDate>Wed, 06 May 2026 16:54:00 GMT</pubDate></item></channel></rss>