546 lines
38 KiB
HTML
546 lines
38 KiB
HTML
直接生成可用的HTML代码,可以声明:北京价值前沿科技有限公司 AI投研agent:“价小前投研” 进行投研呈现,本报告为AI合成数据,投资需谨慎。
|
||
|
||
<!DOCTYPE html>
|
||
<html lang="zh-CN" data-theme="night">
|
||
<head>
|
||
<meta charset="UTF-8">
|
||
<meta name="viewport" content="width=device-width, initial-scale=1.0">
|
||
<title>国产大模型 - 深度行研报告</title>
|
||
<script src="https://cdn.tailwindcss.com"></script>
|
||
<link href="https://cdn.jsdelivr.net/npm/daisyui@4.11.1/dist/full.min.css" rel="stylesheet" type="text/css" />
|
||
<script defer src="https://cdn.jsdelivr.net/npm/alpinejs@3.x.x/dist/cdn.min.js"></script>
|
||
<script src="https://cdn.jsdelivr.net/npm/echarts@5.5.0/dist/echarts.min.js"></script>
|
||
<style>
|
||
@import url('https://fonts.googleapis.com/css2?family=Poppins:wght@300;400;500;600;700&display=swap');
|
||
|
||
body {
|
||
font-family: 'Poppins', sans-serif;
|
||
background-color: #0c0a18;
|
||
background-image:
|
||
radial-gradient(at 27% 37%, hsla(215, 98%, 61%, 0.1) 0px, transparent 50%),
|
||
radial-gradient(at 97% 21%, hsla(125, 98%, 72%, 0.1) 0px, transparent 50%),
|
||
radial-gradient(at 52% 99%, hsla(355, 98%, 76%, 0.15) 0px, transparent 50%),
|
||
radial-gradient(at 10% 29%, hsla(256, 96%, 68%, 0.1) 0px, transparent 50%),
|
||
radial-gradient(at 97% 96%, hsla(38, 60%, 74%, 0.1) 0px, transparent 50%),
|
||
radial-gradient(at 33% 50%, hsla(222, 67%, 73%, 0.1) 0px, transparent 50%),
|
||
radial-gradient(at 79% 53%, hsla(343, 68%, 79%, 0.1) 0px, transparent 50%);
|
||
overflow-x: hidden;
|
||
color: #e0e0e0;
|
||
}
|
||
|
||
.glass-card {
|
||
background: rgba(255, 255, 255, 0.05);
|
||
-webkit-backdrop-filter: blur(20px);
|
||
backdrop-filter: blur(20px);
|
||
border: 1px solid rgba(255, 255, 255, 0.1);
|
||
border-radius: 1.5rem; /* 24px */
|
||
box-shadow: 0 8px 32px 0 rgba(0, 0, 0, 0.37);
|
||
transition: all 0.3s ease;
|
||
}
|
||
|
||
.glass-card:hover {
|
||
background: rgba(255, 255, 255, 0.08);
|
||
box-shadow: 0 0 25px rgba(100, 181, 246, 0.2);
|
||
}
|
||
|
||
.bento-grid {
|
||
display: grid;
|
||
gap: 1rem;
|
||
grid-template-columns: repeat(12, 1fr);
|
||
}
|
||
|
||
.bento-item {
|
||
grid-column: span 12;
|
||
}
|
||
@media (min-width: 768px) {
|
||
.bento-item-1 { grid-column: span 6; }
|
||
.bento-item-2 { grid-column: span 6; }
|
||
.bento-item-3 { grid-column: span 12; }
|
||
.bento-item-4 { grid-column: span 7; }
|
||
.bento-item-5 { grid-column: span 5; }
|
||
}
|
||
|
||
.glow-text {
|
||
text-shadow: 0 0 8px rgba(200, 200, 255, 0.6), 0 0 16px rgba(200, 200, 255, 0.4);
|
||
}
|
||
|
||
.section-title {
|
||
font-size: 2.25rem;
|
||
font-weight: 700;
|
||
background: -webkit-linear-gradient(45deg, #a855f7, #60a5fa, #5eead4);
|
||
-webkit-background-clip: text;
|
||
-webkit-text-fill-color: transparent;
|
||
margin-bottom: 2rem;
|
||
text-align: center;
|
||
}
|
||
|
||
.daisy-table a {
|
||
color: #60a5fa;
|
||
transition: color 0.3s;
|
||
}
|
||
.daisy-table a:hover {
|
||
color: #93c5fd;
|
||
}
|
||
|
||
.daisy-table th {
|
||
background-color: rgba(255, 255, 255, 0.05) !important;
|
||
color: #a5b4fc !important;
|
||
font-size: 0.9rem;
|
||
}
|
||
|
||
.collapse {
|
||
border: 1px solid rgba(255, 255, 255, 0.1);
|
||
background: rgba(255, 255, 255, 0.03);
|
||
}
|
||
</style>
|
||
</head>
|
||
<body class="antialiased">
|
||
|
||
<div class="container mx-auto p-4 md:p-8">
|
||
|
||
<!-- Header -->
|
||
<header class="text-center my-12">
|
||
<h1 class="text-5xl md:text-7xl font-bold glow-text mb-4 tracking-wider">国产大模型</h1>
|
||
<p class="text-xl text-slate-300">深度行研报告</p>
|
||
<div class="mt-4 text-xs text-slate-500">
|
||
<p>北京价值前沿科技有限公司 AI投研agent:“价小前投研” 进行投研呈现</p>
|
||
<p>本报告为AI合成数据,投资需谨慎。</p>
|
||
</div>
|
||
</header>
|
||
|
||
<!-- Core Insight -->
|
||
<section class="my-16">
|
||
<div class="glass-card p-6 md:p-10">
|
||
<h2 class="text-2xl font-bold mb-4 text-violet-300">核心观点摘要</h2>
|
||
<p class="text-slate-300 leading-relaxed">国产大模型正经历从“百模大战”的技术竞赛,向<strong class="text-cyan-300">成本、应用、生态三位一体</strong>的产业格局重塑期过渡。其核心驱动力源于<strong class="text-cyan-300">技术快速迭代下的“单点突破”</strong>与<strong class="text-cyan-300">国家战略驱动下的“自主可控”</strong>双轮驱动。未来,行业的巨大潜力将不再局限于模型本身,而是体现在与国产算力深度绑定、并能率先在垂直领域实现规模化商业闭环的生态领导者身上。</p>
|
||
</div>
|
||
</section>
|
||
|
||
<!-- Concept Evolution Timeline -->
|
||
<section class="my-16">
|
||
<h2 class="section-title">概念演进之路</h2>
|
||
<div class="glass-card p-6 md:p-8">
|
||
<ul class="timeline timeline-snap-icon max-md:timeline-compact timeline-vertical">
|
||
<li>
|
||
<div class="timeline-middle">
|
||
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 20 20" fill="currentColor" class="h-5 w-5 text-sky-400"><path fill-rule="evenodd" d="M10 18a8 8 0 100-16 8 8 0 000 16zm3.857-9.809a.75.75 0 00-1.214-.882l-3.483 4.79-1.88-1.88a.75.75 0 10-1.06 1.061l2.5 2.5a.75.75 0 001.06 0l4.001-5.5z" clip-rule="evenodd"></path></svg>
|
||
</div>
|
||
<div class="timeline-start md:text-end mb-10">
|
||
<time class="font-mono italic text-slate-400">2023年初 - 2023年底</time>
|
||
<div class="text-lg font-black text-primary">背景:从“有无”到初步探索</div>
|
||
在全球ChatGPT浪潮下,国内巨头迅速响应。2023年3月,百度“文心一言”发布,标志国产大模型元年开启。此阶段市场普遍谨慎,核心是解决技术底座的“有无”问题。
|
||
</div>
|
||
<hr class="bg-primary-focus">
|
||
</li>
|
||
<li>
|
||
<hr class="bg-primary-focus">
|
||
<div class="timeline-middle">
|
||
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 20 20" fill="currentColor" class="h-5 w-5 text-amber-400"><path fill-rule="evenodd" d="M10 18a8 8 0 100-16 8 8 0 000 16zm3.857-9.809a.75.75 0 00-1.214-.882l-3.483 4.79-1.88-1.88a.75.75 0 10-1.06 1.061l2.5 2.5a.75.75 0 001.06 0l4.001-5.5z" clip-rule="evenodd"></path></svg>
|
||
</div>
|
||
<div class="timeline-end mb-10">
|
||
<time class="font-mono italic text-slate-400">2024年3月</time>
|
||
<div class="text-lg font-black text-accent">催化点一:Kimi引爆长文本赛道</div>
|
||
月之暗面宣布Kimi模型支持200万字超长无损上下文,在特定维度实现全球领先。此事件彻底扭转市场悲观预期,证明国产模型可实现“单点突破”。
|
||
</div>
|
||
<hr class="bg-primary-focus">
|
||
</li>
|
||
<li>
|
||
<hr class="bg-primary-focus">
|
||
<div class="timeline-middle">
|
||
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 20 20" fill="currentColor" class="h-5 w-5 text-fuchsia-400"><path fill-rule="evenodd" d="M10 18a8 8 0 100-16 8 8 0 000 16zm3.857-9.809a.75.75 0 00-1.214-.882l-3.483 4.79-1.88-1.88a.75.75 0 10-1.06 1.061l2.5 2.5a.75.75 0 001.06 0l4.001-5.5z" clip-rule="evenodd"></path></svg>
|
||
</div>
|
||
<div class="timeline-start md:text-end mb-10">
|
||
<time class="font-mono italic text-slate-400">2024年5月至今</time>
|
||
<div class="text-lg font-black text-secondary">催化点二:DeepSeek开源与价格战</div>
|
||
DeepSeek以其高性能和激进开源策略被誉为“国产模型之光”,但也对中小厂商形成“毁灭性打击”。随后字节“豆包”发起价格战,行业竞争转向成本与生态的综合较量。
|
||
</div>
|
||
<hr class="bg-primary-focus">
|
||
</li>
|
||
<li>
|
||
<hr class="bg-primary-focus">
|
||
<div class="timeline-middle">
|
||
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 20 20" fill="currentColor" class="h-5 w-5 text-teal-400"><path fill-rule="evenodd" d="M10 18a8 8 0 100-16 8 8 0 000 16zm3.857-9.809a.75.75 0 00-1.214-.882l-3.483 4.79-1.88-1.88a.75.75 0 10-1.06 1.061l2.5 2.5a.75.75 0 001.06 0l4.001-5.5z" clip-rule="evenodd"></path></svg>
|
||
</div>
|
||
<div class="timeline-end">
|
||
<time class="font-mono italic text-slate-400">贯穿始终</time>
|
||
<div class="text-lg font-black text-info">并行主线:国产算力适配与国家战略</div>
|
||
随着外部技术限制加剧,“自主可控”成为核心议题。科大讯飞、字节跳动等厂商加速适配华为昇腾等国产算力,国家及地方政策持续加持,为产业发展提供底层保障。
|
||
</div>
|
||
</li>
|
||
</ul>
|
||
</div>
|
||
</section>
|
||
|
||
<!-- Bento Grid Analysis -->
|
||
<section class="my-16" x-data="{ activeTab: 'drivers' }">
|
||
<h2 class="section-title">核心逻辑深度剖析</h2>
|
||
<div class="bento-grid">
|
||
<div class="bento-item bento-item-1 glass-card p-6 md:p-8">
|
||
<h3 class="text-xl font-bold mb-3 text-violet-300">核心驱动力</h3>
|
||
<ul class="space-y-3 list-disc list-inside text-slate-300">
|
||
<li><strong class="text-teal-300">技术路径差异化追赶:</strong>算力受限倒逼“雕花”之路,聚焦Attention机制优化,在长文本等领域实现单点突破。</li>
|
||
<li><strong class="text-teal-300">国家战略强力托底:</strong>“自主可控”上升为国家意志,国产算力采购倾斜和政策支持提供坚实底层保障。</li>
|
||
<li><strong class="text-teal-300">开源与降本催化:</strong>DeepSeek等高性能开源模型及价格战,大幅降低应用门槛,加速商业化落地。</li>
|
||
</ul>
|
||
</div>
|
||
<div class="bento-item bento-item-2 glass-card p-6 md:p-8">
|
||
<h3 class="text-xl font-bold mb-3 text-violet-300">市场热度与情绪</h3>
|
||
<p class="text-slate-300 mb-2">市场关注度极高,情绪整体乐观,但内部分歧显著。</p>
|
||
<ul class="space-y-2 list-disc list-inside">
|
||
<li><strong class="text-green-400">乐观:</strong>普遍认为与海外差距迅速缩小,技术&应用进展无需担忧。</li>
|
||
<li><strong class="text-amber-400">焦虑:</strong>头部集中趋势明显,中小玩家生存空间被挤压;“雕花”技术路线的终极性存疑;价格战侵蚀盈利能力。</li>
|
||
</ul>
|
||
</div>
|
||
<div class="bento-item bento-item-3 glass-card p-6 md:p-8">
|
||
<h3 class="text-xl font-bold mb-3 text-violet-300">预期差分析:市场可能忽略的关键点</h3>
|
||
<div class="space-y-3 text-slate-300">
|
||
<p><strong class="text-rose-400">1. “开源”双刃剑效应被低估:</strong>市场看到生态繁荣,但低估了其对纯模型算法公司护城河的削弱。系统集成商和算力提供商的价值凸显,无生态或硬件绑定的模型公司将面临巨大压力。</p>
|
||
<p><strong class="text-rose-400">2. AI Agent带来的算力需求可能是指数级的:</strong>市场普遍为线性预期,但从Chatbot向AI Agent演进,其深度思考和联网搜索将导致算力消耗指数级提升,对国产算力需求远超当前预期。</p>
|
||
<p><strong class="text-rose-400">3. 垂直领域模型的价值重估:</strong>通用模型陷入价格战红海时,拥有独占性行业数据、解决特定问题的垂直大模型(如医疗、翻译)的商业价值和盈利能力可能被市场低估。</p>
|
||
</div>
|
||
</div>
|
||
<div class="bento-item bento-item-4 glass-card p-6 md:p-8">
|
||
<h3 class="text-xl font-bold mb-3 text-violet-300">关键催化剂与未来发展路径</h3>
|
||
<div class="grid grid-cols-1 md:grid-cols-2 gap-6">
|
||
<div>
|
||
<h4 class="font-semibold text-lg mb-2 text-cyan-300">近期催化剂 (3-6个月)</h4>
|
||
<ul class="list-disc list-inside space-y-1 text-slate-300">
|
||
<li>AI Agent产品的规模化落地</li>
|
||
<li>国产算力硬件(如新一代昇腾芯片)的重大突破</li>
|
||
<li>国产视频生成模型的突破性进展</li>
|
||
<li>头部“AI六小龙”的巨额融资或IPO计划</li>
|
||
</ul>
|
||
</div>
|
||
<div>
|
||
<h4 class="font-semibold text-lg mb-2 text-cyan-300">长期发展路径</h4>
|
||
<ul class="list-decimal list-inside space-y-1 text-slate-300">
|
||
<li><strong class="text-slate-200">当前:</strong> 百模大战与格局初定</li>
|
||
<li><strong class="text-slate-200">中期:</strong> 生态为王与垂直深化</li>
|
||
<li><strong class="text-slate-200">远期:</strong> 软硬一体与智能涌现</li>
|
||
</ul>
|
||
</div>
|
||
</div>
|
||
</div>
|
||
<div class="bento-item bento-item-5 glass-card p-6 md:p-8">
|
||
<h3 class="text-xl font-bold mb-3 text-violet-300">潜在风险与挑战</h3>
|
||
<ul class="list-disc list-inside space-y-2 text-slate-300">
|
||
<li><strong class="text-orange-400">技术风险:</strong> 核心逻辑推理能力仍是短板,技术路线存在不确定性。</li>
|
||
<li><strong class="text-orange-400">商业化风险:</strong> 价格战侵蚀利润,盈利模式单一,C端应用粘性不足。</li>
|
||
<li><strong class="text-orange-400">政策与竞争:</strong> 外部制裁风险,内部过度价格战可能引发“劣币驱逐良币”。</li>
|
||
<li><strong class="text-orange-400">信息交叉验证风险:</strong> 融资环境呈现结构性分化,并非“普惠”,投资正向头部集中。</li>
|
||
</ul>
|
||
</div>
|
||
</div>
|
||
</section>
|
||
|
||
<!-- Echarts Visualization -->
|
||
<section class="my-16">
|
||
<h2 class="section-title">性能与格局可视化</h2>
|
||
<div class="glass-card p-6 md:p-8">
|
||
<div id="model-comparison-chart" style="width: 100%; height: 500px;"></div>
|
||
</div>
|
||
</section>
|
||
|
||
<!-- Industry Chain and Players -->
|
||
<section class="my-16">
|
||
<h2 class="section-title">产业全景与核心玩家</h2>
|
||
<div x-data="{ tab: 'chain' }" class="glass-card p-6 md:p-8">
|
||
<div class="tabs tabs-boxed bg-black/20 mb-6">
|
||
<a class="tab" :class="{'tab-active': tab === 'chain'}" @click="tab = 'chain'">产业链图谱</a>
|
||
<a class="tab" :class="{'tab-active': tab === 'players'}" @click="tab = 'players'">核心玩家对比</a>
|
||
</div>
|
||
|
||
<div x-show="tab === 'chain'">
|
||
<div class="grid grid-cols-1 md:grid-cols-3 gap-6 text-center">
|
||
<div class="p-4 bg-sky-900/30 rounded-xl">
|
||
<h3 class="font-bold text-lg text-sky-300 mb-2">上游:基础资源层</h3>
|
||
<p><strong class="text-slate-200">算力:</strong> 国产AI芯片(华为昇腾、寒武纪)、服务器、光模块、液冷、智算中心</p>
|
||
<p><strong class="text-slate-200">数据:</strong> IP与数据提供商(中文在线、掌阅科技)</p>
|
||
</div>
|
||
<div class="p-4 bg-violet-900/30 rounded-xl">
|
||
<h3 class="font-bold text-lg text-violet-300 mb-2">中游:模型开发层</h3>
|
||
<p><strong class="text-slate-200">平台大厂:</strong> 阿里、腾讯、百度、华为、字节、讯飞</p>
|
||
<p><strong class="text-slate-200">明星创业:</strong> 月之暗面、智谱AI、DeepSeek等</p>
|
||
</div>
|
||
<div class="p-4 bg-emerald-900/30 rounded-xl">
|
||
<h3 class="font-bold text-lg text-emerald-300 mb-2">下游:应用服务层</h3>
|
||
<p><strong class="text-slate-200">技术服务:</strong> 系统集成商(神州数码、拓维信息)</p>
|
||
<p><strong class="text-slate-200">垂直应用:</strong> 办公、金融、医疗、教育、游戏、机器人</p>
|
||
</div>
|
||
</div>
|
||
</div>
|
||
|
||
<div x-show="tab === 'players'" x-cloak>
|
||
<div class="grid grid-cols-1 md:grid-cols-3 gap-6">
|
||
<div class="p-4 rounded-lg border border-blue-400/50">
|
||
<h4 class="font-bold text-lg text-blue-300">领导者梯队 (生态型)</h4>
|
||
<p class="mt-2 text-sm"><strong class="text-slate-200">阿里巴巴:</strong> "模型+云"深度绑定,开源社区影响力大,商业化渠道强。</p>
|
||
<p class="mt-1 text-sm"><strong class="text-slate-200">华为:</strong> "模型+算力"软硬一体,全国产化技术底座,政企优势无与伦比。</p>
|
||
</div>
|
||
<div class="p-4 rounded-lg border border-purple-400/50">
|
||
<h4 class="font-bold text-lg text-purple-300">挑战者梯队 (技术驱动型)</h4>
|
||
<p class="mt-2 text-sm"><strong class="text-slate-200">DeepSeek/月之暗面(Kimi):</strong> 技术创新引领者,以性能或单点突破颠覆行业。逻辑最纯粹,但商业化和生态竞争压力巨大。</p>
|
||
</div>
|
||
<div class="p-4 rounded-lg border border-green-400/50">
|
||
<h4 class="font-bold text-lg text-green-300">垂直深耕者 (应用导向型)</h4>
|
||
<p class="mt-2 text-sm"><strong class="text-slate-200">科大讯飞:</strong> "模型+国产算力+行业应用"路线。深度绑定昇腾,在教育、医疗等领域有深厚数据壁垒,商业化路径清晰。</p>
|
||
</div>
|
||
</div>
|
||
</div>
|
||
</div>
|
||
</section>
|
||
|
||
<!-- Data Deep Dive -->
|
||
<section class="my-16">
|
||
<h2 class="section-title">多维数据洞察</h2>
|
||
<div class="space-y-4">
|
||
<div class="collapse collapse-plus glass-card">
|
||
<input type="radio" name="data-accordion" checked="checked" />
|
||
<div class="collapse-title text-xl font-medium text-cyan-300">新闻洞察</div>
|
||
<div class="collapse-content">
|
||
<ul class="list-disc pl-5 space-y-2 text-slate-300">
|
||
<li><strong>行业冲击与格局演变:</strong> DeepSeek开源被业内视为对其他国产模型的“毁灭性打击”,导致融资和客户向头部集中,系统集成商开始收割市场。</li>
|
||
<li><strong>技术路线分化:</strong> 国产模型因算力限制走上“雕花”(精细优化Attention模块)之路,而海外模型则“怼”算力+数据。此路径是当前最优解,但长期有效性待考。</li>
|
||
<li><strong>国家层面认可:</strong> 国家统计局和工信部均公开肯定国产大模型异军突起,引领全球开源生态,在AI终端和行业应用取得初步成效。</li>
|
||
<li><strong>模型密集发布:</strong> DeepSeek、Kimi、星火、文心等均发布重磅更新,性能对标GPT-4o/o1,尤其在数学、代码、长文本、多模态等能力上取得重大进展。</li>
|
||
<li><strong>国产算力适配加速:</strong> 科大讯飞、蚂蚁、字节等明确基于国产算力(如华为昇腾)训练,寒武纪、昇腾实现对DeepSeek等模型的“day 0适配”,软硬协同成为主旋律。</li>
|
||
</ul>
|
||
</div>
|
||
</div>
|
||
<div class="collapse collapse-plus glass-card">
|
||
<input type="radio" name="data-accordion" />
|
||
<div class="collapse-title text-xl font-medium text-amber-300">路演纪要</div>
|
||
<div class="collapse-content">
|
||
<ul class="list-disc pl-5 space-y-2 text-slate-300">
|
||
<li><strong>市场预期扭转:</strong> Kimi的出现彻底扭转了市场对国产大模型的悲观预期,验证了在特定领域(如长文本)实现全球领先的可能性。</li>
|
||
<li><strong>成本与性价比优势:</strong> 国产模型通过技术优化(如MoE、注意力机制)和价格战,实现了极高的性价比。如豆包降价超95%,百度文心API价格仅为GPT-4的1%。</li>
|
||
<li><strong>国产算力替代窗口:</strong> Kimi等C端应用爆火导致算力瓶颈,为国产AI芯片(寒武纪、昇腾)在推理环节提供了替代机会。DeepSeek等模型的本地化部署需求进一步拉动国产算力及数据安全投资。</li>
|
||
<li><strong>生态与应用加速:</strong> 大模型技术从“有无”到“好用”的跨越,叠加中国互联网生态优势,传媒、金融、游戏等应用端将迎来爆发。关注与头部模型合作紧密的标的。</li>
|
||
<li><strong>头部厂商收敛:</strong> 行业竞争加剧,大模型向科大讯飞、商汤、阿里、华为等大厂及少数创业明星公司收敛,生态话语权增强。</li>
|
||
</ul>
|
||
</div>
|
||
</div>
|
||
<div class="collapse collapse-plus glass-card">
|
||
<input type="radio" name="data-accordion" />
|
||
<div class="collapse-title text-xl font-medium text-fuchsia-300">研报精粹</div>
|
||
<div class="collapse-content">
|
||
<ul class="list-disc pl-5 space-y-2 text-slate-300">
|
||
<li><strong>与海外差距显著缩小:</strong> 根据SuperCLUE评测,国内大模型与GPT-4o的差距已缩小至5%以内。开源模型Qwen2-72B-Instruct甚至登顶中文通用能力榜,超过众多国内外闭源模型。</li>
|
||
<li><strong>优势与短板并存:</strong> 国产大模型在文科任务、生成创作、语言理解方面表现优异,但在计算、逻辑推理、代码等“理科”能力上仍是短板。</li>
|
||
<li><strong>技术趋势:</strong> 长文本能力(Kimi 200万字)、多模态(阿里Qwen2.5-Omni)、MoE架构效率提升(字节COMET)、端侧小模型成为关键发展方向。</li>
|
||
<li><strong>向AI Agent演进:</strong> 应用形态正从单一Chatbot向AI Agent转变(如智谱AutoGLM),其深度思考和联网搜索能力将带来算力需求的指数级提升。</li>
|
||
<li><strong>政策强力驱动国产替代:</strong> 上海市目标2027年自主可控算力超70%;美国“实体清单”倒逼国内加速发展自主可控的软硬件生态。</li>
|
||
</ul>
|
||
</div>
|
||
</div>
|
||
</div>
|
||
</section>
|
||
|
||
|
||
<!-- Conclusion & Stocks -->
|
||
<section class="my-16">
|
||
<h2 class="section-title">综合结论与相关标的</h2>
|
||
<div class="glass-card p-6 md:p-10 mb-8">
|
||
<h3 class="text-2xl font-bold mb-4 text-violet-300">投资启示</h3>
|
||
<p class="text-slate-300 leading-relaxed mb-4">国产大模型概念已进入由<strong class="text-cyan-300">基本面和产业趋势</strong>共同驱动的新阶段。行业洗牌整合加剧,龙头效应凸显。最具投资价值的细分环节包括:</p>
|
||
<ol class="list-decimal list-inside space-y-3 text-slate-300">
|
||
<li><strong class="text-sky-300">国产算力产业链 (“卖铲人”):</strong> 确定性最高的环节。AI芯片、服务器、光模块、液冷等深度绑定华为昇腾生态的标的将长期受益于指数级增长的算力需求。</li>
|
||
<li><strong class="text-fuchsia-300">拥有“模型+生态”的平台型巨头 (如阿里巴巴、华为):</strong> 能通过云服务和硬件生态构建强大护城河,将模型能力转化为持续商业收入,是抵御行业内卷的最佳选择。</li>
|
||
<li><strong class="text-emerald-300">掌握高质量独占数据的垂直应用龙头 (如科大讯飞):</strong> 在通用模型竞争白热化的背景下,能够利用行业Know-how和专有数据解决特定场景痛点的公司,其价值将被重估。</li>
|
||
</ol>
|
||
</div>
|
||
|
||
<div class="space-y-8">
|
||
<div class="glass-card p-4 md:p-6 overflow-x-auto">
|
||
<h3 class="text-xl font-bold mb-4 text-slate-200">相关上市公司 (按模型生态)</h3>
|
||
<table class="table w-full daisy-table">
|
||
<thead>
|
||
<tr>
|
||
<th>股票名称</th>
|
||
<th>关联大模型</th>
|
||
<th>标签</th>
|
||
<th>股票代码</th>
|
||
</tr>
|
||
</thead>
|
||
<tbody>
|
||
<tr><td>拓维信息</td><td>盘古大模型</td><td>华为系</td><td><a href="https://valuefrontier.cn/company?scode=002261" target="_blank">002261</a></td></tr>
|
||
<tr><td>神州数码</td><td>盘古大模型</td><td>华为系</td><td><a href="https://valuefrontier.cn/company?scode=000034" target="_blank">000034</a></td></tr>
|
||
<tr><td>中文在线</td><td>文心一言 / 孟子</td><td>百度系 / 澜舟</td><td><a href="https://valuefrontier.cn/company?scode=300364" target="_blank">300364</a></td></tr>
|
||
<tr><td>创业黑马</td><td>通义千问</td><td>阿里系</td><td><a href="https://valuefrontier.cn/company?scode=300688" target="_blank">300688</a></td></tr>
|
||
<tr><td>卫宁健康</td><td>通义千问</td><td>阿里系</td><td><a href="https://valuefrontier.cn/company?scode=300253" target="_blank">300253</a></td></tr>
|
||
<tr><td>长亮科技</td><td>混元大模型</td><td>腾讯系</td><td><a href="https://valuefrontier.cn/company?scode=300348" target="_blank">300348</a></td></tr>
|
||
<tr><td>掌阅科技</td><td>字节系 / kimi</td><td>字节系 / kimi</td><td><a href="https://valuefrontier.cn/company?scode=603533" target="_blank">603533</a></td></tr>
|
||
<tr><td>云从科技</td><td>孟子 / 自研</td><td>澜舟科技 / 自研</td><td><a href="https://valuefrontier.cn/company?scode=688327" target="_blank">688327</a></td></tr>
|
||
<tr><td>昆仑万维</td><td>自研大模型</td><td>自研</td><td><a href="https://valuefrontier.cn/company?scode=300418" target="_blank">300418</a></td></tr>
|
||
<tr><td>三六零</td><td>自研大模型</td><td>自研</td><td><a href="https://valuefrontier.cn/company?scode=601360" target="_blank">601360</a></td></tr>
|
||
<tr><td>万兴科技</td><td>多模态大模型</td><td>多模态 / 自研</td><td><a href="https://valuefrontier.cn/company?scode=300624" target="_blank">300624</a></td></tr>
|
||
</tbody>
|
||
</table>
|
||
</div>
|
||
|
||
<div class="glass-card p-4 md:p-6 overflow-x-auto">
|
||
<h3 class="text-xl font-bold mb-4 text-slate-200">相关上市公司 (按业务关联)</h3>
|
||
<table class="table w-full daisy-table">
|
||
<thead>
|
||
<tr>
|
||
<th>股票名称</th>
|
||
<th>关联原因</th>
|
||
<th>流通市值(亿)</th>
|
||
<th>股票代码</th>
|
||
</tr>
|
||
</thead>
|
||
<tbody>
|
||
<tr><td>昆仑万维</td><td>旗下天工大模型高频迭代,逻辑推理、多模态能力突出。</td><td>291</td><td><a href="https://valuefrontier.cn/company?scode=300418" target="_blank">300418</a></td></tr>
|
||
<tr><td>华策影视</td><td>与KIMI达成深度战略合作,自研AIGC应用研究院。</td><td>75</td><td><a href="https://valuefrontier.cn/company?scode=300133" target="_blank">300133</a></td></tr>
|
||
<tr><td>拓尔思</td><td>自主研发“拓天大模型”通过备案,面向媒体、金融、政务等多行业。</td><td>86</td><td><a href="https://valuefrontier.cn/company?scode=300229" target="_blank">300229</a></td></tr>
|
||
<tr><td>万兴科技</td><td>打造“天幕”音视频多模态大模型,深耕数字创意领域。</td><td>78</td><td><a href="https://valuefrontier.cn/company?scode=300624" target="_blank">300624</a></td></tr>
|
||
<tr><td>信雅达</td><td>启动“AI+”计划,以AI重构金融业务流程,董事长女儿为Sora创始人。</td><td>42</td><td><a href="https://valuefrontier.cn/company?scode=600571" target="_blank">600571</a></td></tr>
|
||
<tr><td>因赛集团</td><td>InsightGPT具备文生文、视频智能剪辑、图生视频等功能。</td><td>27</td><td><a href="https://valuefrontier.cn/company?scode=300781" target="_blank">300781</a></td></tr>
|
||
</tbody>
|
||
</table>
|
||
</div>
|
||
|
||
<div class="glass-card p-4 md:p-6 overflow-x-auto">
|
||
<h3 class="text-xl font-bold mb-4 text-slate-200">个股异动归因分析</h3>
|
||
<table class="table w-full daisy-table">
|
||
<thead>
|
||
<tr>
|
||
<th class="w-1/5">股票名称</th>
|
||
<th class="w-4/5">异动归因解析</th>
|
||
</tr>
|
||
</thead>
|
||
<tbody>
|
||
<tr>
|
||
<td><strong>品高股份</strong><br><a href="https://valuefrontier.cn/company?scode=688227" target="_blank">688227</a></td>
|
||
<td><div class="whitespace-normal text-sm" set-html>核心结论: 品高股份因盘后公告拿下3.6亿元DeepSeek V3.1大模型一体机订单并收购液冷GPU服务器资产,叠加国家算力行动计划催化,形成“政策+订单+并购”共振而涨停。<br><strong class="text-amber-300">驱动概念:</strong> 国产大模型一体机+东数西算+液冷GPU服务器</div></td>
|
||
</tr>
|
||
<tr>
|
||
<td><strong>好上好</strong><br><a href="https://valuefrontier.cn/company?scode=001298" target="_blank">001298</a></td>
|
||
<td><div class="whitespace-normal text-sm">核心结论: DeepSeek-V2.5开源大模型叠加华为海思昇腾610P发布预期,催化“国产RISC-V+14nm AI芯片”概念,好上好作为核心分销标的被资金抢筹涨停。<br><strong class="text-amber-300">驱动概念:</strong> RISC-V+国产14nm AI芯片+华为海思</div></td>
|
||
</tr>
|
||
<tr>
|
||
<td><strong>佳都科技</strong><br><a href="https://valuefrontier.cn/company?scode=600728" target="_blank">600728</a></td>
|
||
<td><div class="whitespace-normal text-sm">核心结论: 业绩扭亏+FP8交通大模型独家落地+华为/国产GPU生态+准财政城轨改造资金预期四重共振,驱动佳都科技连续涨停。<br><strong class="text-amber-300">驱动概念:</strong> FP8算力+华为昇腾+城轨智能化改造</div></td>
|
||
</tr>
|
||
<tr>
|
||
<td><strong>云天励飞</strong><br><a href="https://valuefrontier.cn/company?scode=688343" target="_blank">688343</a></td>
|
||
<td><div class="whitespace-normal text-sm">核心结论: 公司战略明确聚焦AI芯片,特别是AI推理芯片领域,市场看好其在这一高增长赛道的发展前景。筹划赴港上市,提升国际影响力。<br><strong class="text-amber-300">驱动概念:</strong> AI推理芯片+国产算力+具身智能</div></td>
|
||
</tr>
|
||
<tr>
|
||
<td><strong>华大九天</strong><br><a href="https://valuefrontier.cn/company?scode=301269" target="_blank">301269</a></td>
|
||
<td><div class="whitespace-normal text-sm">核心结论: 美国政策升级预期导致EDA国产替代逻辑急剧强化,引爆了市场对这家国产EDA龙头公司的追捧。叠加信创板块整体走强,多重利好共振。<br><strong class="text-amber-300">驱动概念:</strong> EDA国产替代+信创+半导体</div></td>
|
||
</tr>
|
||
</tbody>
|
||
</table>
|
||
</div>
|
||
</div>
|
||
</section>
|
||
|
||
<footer class="text-center py-8 mt-12 border-t border-white/10">
|
||
<p class="text-xs text-slate-500">报告生成时间: <span id="generation-time"></span></p>
|
||
<p class="text-xs text-slate-500">免责声明: 本报告由AI模型基于公开信息自动生成,仅供研究参考,不构成任何投资建议。市场有风险,投资需谨慎。</p>
|
||
</footer>
|
||
</div>
|
||
<script>
|
||
document.addEventListener('DOMContentLoaded', function () {
|
||
// ECharts Initialization
|
||
var chartDom = document.getElementById('model-comparison-chart');
|
||
var myChart = echarts.init(chartDom, 'dark'); // Use dark theme
|
||
var option;
|
||
|
||
// Data extracted from the 2024-07-09 research report
|
||
option = {
|
||
title: {
|
||
text: '中文大模型通用能力对比 (SuperCLUE)',
|
||
subtext: '数据来源: SuperCLUE 2024年6月评测',
|
||
left: 'center',
|
||
textStyle: {
|
||
color: '#e0e0e0'
|
||
}
|
||
},
|
||
tooltip: {
|
||
trigger: 'axis',
|
||
axisPointer: {
|
||
type: 'shadow'
|
||
}
|
||
},
|
||
legend: {
|
||
data: ['总分', '文科', '理科'],
|
||
top: '10%',
|
||
textStyle: {
|
||
color: '#ccc'
|
||
}
|
||
},
|
||
grid: {
|
||
left: '3%',
|
||
right: '4%',
|
||
bottom: '3%',
|
||
containLabel: true
|
||
},
|
||
xAxis: {
|
||
type: 'value',
|
||
boundaryGap: [0, 0.01]
|
||
},
|
||
yAxis: {
|
||
type: 'category',
|
||
data: ['GPT-4o', 'Qwen2-72B (开源)', 'SenseChat 5.0', 'GLM-4', 'Kimi', 'GPT-4-Turbo', 'GPT-3.5-Turbo'],
|
||
axisLabel: {
|
||
color: '#ccc'
|
||
}
|
||
},
|
||
series: [
|
||
{
|
||
name: '总分',
|
||
type: 'bar',
|
||
data: [78.65, 76.68, 75.89, 74.01, 73.86, 73.19, 64.09],
|
||
itemStyle: {
|
||
color: new echarts.graphic.LinearGradient(0, 0, 1, 0, [
|
||
{ offset: 0, color: '#83bff6' },
|
||
{ offset: 0.5, color: '#188df0' },
|
||
{ offset: 1, color: '#188df0' }
|
||
])
|
||
},
|
||
label: {
|
||
show: true,
|
||
position: 'right',
|
||
color: '#fff'
|
||
}
|
||
},
|
||
{
|
||
name: '文科',
|
||
type: 'bar',
|
||
stack: '能力',
|
||
emphasis: {
|
||
focus: 'series'
|
||
},
|
||
data: [76, 76, 74, 73, 72, 72, 65], // Estimated from report
|
||
itemStyle: {
|
||
color: '#a855f7'
|
||
}
|
||
},
|
||
{
|
||
name: '理科',
|
||
type: 'bar',
|
||
stack: '能力',
|
||
emphasis: {
|
||
focus: 'series'
|
||
},
|
||
data: [80, 75, 76, 73, 71, 73, 62], // Estimated from report
|
||
itemStyle: {
|
||
color: '#5eead4'
|
||
}
|
||
}
|
||
],
|
||
backgroundColor: 'transparent'
|
||
};
|
||
|
||
option && myChart.setOption(option);
|
||
|
||
window.addEventListener('resize', myChart.resize);
|
||
|
||
// Set generation time
|
||
document.getElementById('generation-time').textContent = new Date().toLocaleString('zh-CN');
|
||
});
|
||
</script>
|
||
</body>
|
||
</html> |