<?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[LLM API费效比图（每个任务的平均花费，cost per task）]]></title><description><![CDATA[<p dir="auto">原文章链接：<a href="https://artificialanalysis.ai/articles/glm-5-2-is-the-new-leading-open-weights-model-on-the-artificial-analysis-intelligence-index" rel="nofollow ugc">https://artificialanalysis.ai/articles/glm-5-2-is-the-new-leading-open-weights-model-on-the-artificial-analysis-intelligence-index</a></p>
<p dir="auto">这篇文章本来是吹GLM5.2的，但后面的费效比图非常有趣，如下：</p>
<p dir="auto"><img src="https://upload.lcz.me/uploads/8e3c2d11-96b1-4606-a282-983a4aa80387.jpeg" alt="3108f8dc-8283-43e9-b5a7-db43c8f4ade0-image.jpeg" class=" img-fluid img-markdown" /></p>
<p dir="auto">下图是每个模型的“费效比”（横坐标，平均执行benchmark的每个项目的总花费，纵坐标，总体benchmark的评分）</p>
<pre><code>注意，横坐标价格是对数坐标，右侧模型的花费是指数级增长的。
</code></pre>
<p dir="auto">总结来看，Mimo-v2.5 Pro、DeepSeek-V4系列，用比其他模型少一、甚至两个数量级的钱，就可以跑完相同分数的benchmark。</p>
<p dir="auto">相较于DeepSeek-V4-Pro（MAX思考长度）</p>
<table class="table table-bordered table-striped">
<thead>
<tr>
<th>模型名称</th>
<th>提高分数比例</th>
<th>消费提高比例</th>
</tr>
</thead>
<tbody>
<tr>
<td>GLM5.2</td>
<td>15.9%</td>
<td>1000%</td>
</tr>
<tr>
<td>GPT5.5</td>
<td>25%</td>
<td>2000%</td>
</tr>
<tr>
<td>Claude Opus 4.8</td>
<td>27%</td>
<td>4000%</td>
</tr>
<tr>
<td>Claude Fable 5</td>
<td>36%</td>
<td>6000%</td>
</tr>
</tbody>
</table>
]]></description><link>https://lcz.me/topic/602/llm-api费效比图-每个任务的平均花费-cost-per-task</link><generator>RSS for Node</generator><lastBuildDate>Sun, 26 Jul 2026 23:29:53 GMT</lastBuildDate><atom:link href="https://lcz.me/topic/602.rss" rel="self" type="application/rss+xml"/><pubDate>Thu, 18 Jun 2026 01:30:27 GMT</pubDate><ttl>60</ttl><item><title><![CDATA[Reply to LLM API费效比图（每个任务的平均花费，cost per task） on Thu, 18 Jun 2026 12:15:22 GMT]]></title><description><![CDATA[<p dir="auto">目前，性价比最高的还是gpt5.5</p>
]]></description><link>https://lcz.me/post/7338</link><guid isPermaLink="true">https://lcz.me/post/7338</guid><dc:creator><![CDATA[AGI]]></dc:creator><pubDate>Thu, 18 Jun 2026 12:15:22 GMT</pubDate></item><item><title><![CDATA[Reply to LLM API费效比图（每个任务的平均花费，cost per task） on Thu, 18 Jun 2026 12:10:52 GMT]]></title><description><![CDATA[<p dir="auto">老黄：一张卡才卖你十几万，贵吗？</p>
]]></description><link>https://lcz.me/post/7337</link><guid isPermaLink="true">https://lcz.me/post/7337</guid><dc:creator><![CDATA[terry]]></dc:creator><pubDate>Thu, 18 Jun 2026 12:10:52 GMT</pubDate></item><item><title><![CDATA[Reply to LLM API费效比图（每个任务的平均花费，cost per task） on Thu, 18 Jun 2026 11:20:23 GMT]]></title><description><![CDATA[<p dir="auto">我现在的云端主力就是glm5.2+kimi2.7<br />
<img src="https://upload.lcz.me/uploads/3202ce7b-094f-497b-95c5-5aadee755f69.jpeg" alt="c6f77f10-91c0-47fa-8e35-a8debe982990-image.jpeg" class=" img-fluid img-markdown" /><br />
<img src="https://upload.lcz.me/uploads/9e19f0f9-ee5a-4bc8-9bb7-c75c4e943f66.jpeg" alt="525532a4-61fd-41e9-bb50-eda8bd50e0e5-image.jpeg" class=" img-fluid img-markdown" /></p>
]]></description><link>https://lcz.me/post/7331</link><guid isPermaLink="true">https://lcz.me/post/7331</guid><dc:creator><![CDATA[Kk Hh]]></dc:creator><pubDate>Thu, 18 Jun 2026 11:20:23 GMT</pubDate></item><item><title><![CDATA[Reply to LLM API费效比图（每个任务的平均花费，cost per task） on Thu, 18 Jun 2026 03:50:22 GMT]]></title><description><![CDATA[<p dir="auto"><a class="plugin-mentions-user plugin-mentions-a" href="/user/566656661" aria-label="Profile: 566656661">@<bdi>566656661</bdi></a> 主要是民间大神们一起努力去探索,慢慢把资本家的遮羞部全部扯掉</p>
]]></description><link>https://lcz.me/post/7261</link><guid isPermaLink="true">https://lcz.me/post/7261</guid><dc:creator><![CDATA[stxpnet]]></dc:creator><pubDate>Thu, 18 Jun 2026 03:50:22 GMT</pubDate></item><item><title><![CDATA[Reply to LLM API费效比图（每个任务的平均花费，cost per task） on Thu, 18 Jun 2026 03:49:42 GMT]]></title><description><![CDATA[<p dir="auto">皮衣黄: ,好,很好,非常好.<br />
要的就是这种效果!</p>
]]></description><link>https://lcz.me/post/7260</link><guid isPermaLink="true">https://lcz.me/post/7260</guid><dc:creator><![CDATA[stxpnet]]></dc:creator><pubDate>Thu, 18 Jun 2026 03:49:42 GMT</pubDate></item><item><title><![CDATA[Reply to LLM API费效比图（每个任务的平均花费，cost per task） on Thu, 18 Jun 2026 02:04:57 GMT]]></title><description><![CDATA[<p dir="auto"><a class="plugin-mentions-user plugin-mentions-a" href="/user/mark" aria-label="Profile: mark">@<bdi>mark</bdi></a></p>
<p dir="auto">深度思考的能力的確是外國的模型比較好, 但是長久下來會被價格戰打下來</p>
<p dir="auto">倒不如說模型的能力可以依靠原資料堆起來, 但是能Cost down才能看到模型團隊的設計能力</p>
]]></description><link>https://lcz.me/post/7252</link><guid isPermaLink="true">https://lcz.me/post/7252</guid><dc:creator><![CDATA[566656661]]></dc:creator><pubDate>Thu, 18 Jun 2026 02:04:57 GMT</pubDate></item><item><title><![CDATA[Reply to LLM API费效比图（每个任务的平均花费，cost per task） on Thu, 18 Jun 2026 01:54:58 GMT]]></title><description><![CDATA[<p dir="auto">一个比一个贵.这个就是涨价离谱了</p>
]]></description><link>https://lcz.me/post/7246</link><guid isPermaLink="true">https://lcz.me/post/7246</guid><dc:creator><![CDATA[mark]]></dc:creator><pubDate>Thu, 18 Jun 2026 01:54:58 GMT</pubDate></item></channel></rss>