<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>DeepSeek · TopAIMap</title>
    <link>https://topaimap.com/tag/deepseek.html</link>
    <description>深度求索的开放权重模型，以训练成本低出名。 TopAIMap 收录的 DeepSeek 相关简讯，按时间排列。</description>
    <language>zh-CN</language>
    <lastBuildDate>Wed, 14 Oct 2026 08:00:00 +0000</lastBuildDate>
    <atom:link href="https://topaimap.com/tag/deepseek.xml" rel="self" type="application/rss+xml"></atom:link>
    <item>
      <title>DeepSeek V4.1 Flash 入榜 AA-AgentPerf：GB300 达约 2.61 亿 tok/s/MW</title>
      <link>https://topaimap.com/article/deepseek-v41-flash-aa-agentperf.html</link>
      <guid>https://topaimap.com/article/deepseek-v41-flash-aa-agentperf.html</guid>
      <pubDate>Wed, 14 Oct 2026 08:00:00 +0000</pubDate>
      <description>Artificial Analysis 将 DeepSeek V4.1 Flash 加入智能体推理基准 AA-AgentPerf；在每用户 100 tok/s 设定下，NVIDIA GB300 系统吞吐约 261M tokens/s per MW，位列当面硬件榜首。（来源：Artificial Analysis）</description>
      <category>模型</category>
      <category>GPT</category>
      <category>Claude</category>
      <category>Gemini</category>
      <category>DeepSeek</category>
    </item>
  </channel>
</rss>
