100多个AI模型，同一个擂台，用户说了算。
9月16日，B站正式上线AI无限竞技场。这是一个开放的AI模型对战平台：用户可以同时与多个AI模型对话，投票选择回答更好的一方，形成实时排名。目前已有100+个模型参与竞技，覆盖GPT、Claude、Gemini、GLM、通义、文心等主流模型家族。
排行榜：GPT-6登顶，国产占前三两席
上线首日的排行榜数据颇具看点：GPT-6 Astra位居榜首，GLM-5.3排名第二，Claude Fable 5.1排名第三。前五名中国产模型占据了三席。
这个排名的有趣之处在于，它不是技术benchmark的比拼，而是用户体验的直接投票。GPT-6登顶并不意外——OpenAI在通用对话能力上的积累依然领先。但GLM-5.3力压Claude排名第二，说明国产模型在中文场景下的用户体验已经具备了与全球顶级模型竞争的实力。
UP主测评：从专业视角拆解AI
B站同步邀请了10位头部UP主对竞技场进行深度测评。这些UP主覆盖科技、创意、教育等不同领域，从各自的专业视角对AI模型的表现进行拆解。这种"用户生成内容+AI评测"的结合，是B站独有的生态优势——其他平台很难同时拥有活跃的创作者社区和大规模的AI用户群体。
1.9亿月活：B站的AI内容飞轮
竞技场上线的底气来自B站AI内容的爆发式增长。数据显示，B站AI内容消费时长同比增长72%，月均AI内容用户达1.9亿。这意味着B站已经不仅仅是一个"看视频"的平台——它正在成为AI模型的分发渠道和评测入口。
当用户在B站形成了"用哪个AI模型"的消费习惯，这个偏好就会影响他们在其他场景下的选择。B站AI竞技场的真正价值，不在于谁排名第一，而在于1.9亿月活用户的AI使用偏好数据。这些数据对AI厂商来说，比任何技术benchmark都更有商业价值。
明天见。
100+ AI models. One arena. Users decide.
On September 16, Bilibili officially launched the AI Infinite Arena — an open AI model competition platform where users converse with multiple models simultaneously, vote for the better response, and generate real-time rankings. Over 100 models participate, spanning GPT, Claude, Gemini, GLM, Tongyi, and Wenxin model families.
Leaderboard: GPT-6 on Top, Domestic Models Claim 2 of Top 3
Launch-day leaderboard data is telling: GPT-6 Astra leads, GLM-5.3 ranks second, and Claude Fable 5.1 ranks third. Domestic models hold three of the top five spots.
What's interesting: this ranking isn't a technical benchmark competition — it's direct user experience voting. GPT-6 on top isn't surprising — OpenAI's accumulated lead in general conversational ability holds. But GLM-5.3 leapfrogging Claude to second place signals that domestic models' Chinese-language user experience can now compete head-to-head with the world's best.
UP Host Reviews: Expert AI Breakdowns
Bilibili simultaneously invited 10 top UP hosts for in-depth arena reviews. Covering tech, creative, and education verticals, these creators dissect AI model performance from their professional perspectives. This "user-generated content + AI evaluation" combination is Bilibili's unique ecosystem advantage — no other platform simultaneously has an active creator community and a massive AI user base.
190M Monthly Users: Bilibili's AI Content Flywheel
The arena's launch is backed by explosive AI content growth on Bilibili. Data shows AI content watch time up 72% year-over-year, with 190 million monthly AI content users. Bilibili isn't just a "watch videos" platform anymore — it's becoming an AI model distribution channel and evaluation gateway.
When users form "which AI model to use" consumption habits on Bilibili, those preferences carry over to other contexts. The AI Infinite Arena's true value isn't who ranks first — it's the AI usage preference data from 190 million monthly users. For AI companies, that data is more commercially valuable than any technical benchmark.
See you tomorrow.
1.9亿月活用户的AI使用偏好数据，比任何技术benchmark都更有商业价值。
—— Dawn Vision编辑部
AI usage preference data from 190 million monthly users is more commercially valuable than any technical benchmark.
— The Dawn Vision Editorial Desk
Dawn Vision, Bilibili, AI Arena, GPT-6, GLM-5.3, Claude, LLM evaluation, AI content, 190M MAU
Sources · 信源 Sources
本文基于 Dawn Vision 认知引擎处理的 10 个源信号生成，经编辑部人工审核。素材来源：IT之家、B站官方、网易。
This article was generated by the Dawn Vision cognitive engine processing 10 source signals, with human editorial review. Sources: IT之家, Bilibili Official, NetEase.
