单词学习：![[122ef647-d410-4b5d-ace0-19454be8d2c3.mp3]]

新闻播报：![[46806b85-0a12-442f-9899-c43ce23f9134.mp3]]

中国AI模型在OpenRouter平台用量首次超越美国，MiniMax M2.5领跑全球

简报：

• 根据在线AI托管平台OpenRouter的最新数据显示，今年2月，以MiniMax、月之暗面、智谱和DeepSeek为代表的中国开源模型，在全球Token使用量上首次全面超越美国模型。
• 2月9日至15日期间，中国模型的周调用量达到4.12万亿Token，首次超过美国模型的2.94万亿Token。
• 随后一周，中国模型的周调用量进一步增至5.16万亿Token，三周内涨幅高达127%。
• 在2月份的OpenRouter平台排名中，发布仅两周的MiniMax M2.5以4.55万亿Token的调用量位居全球榜首。
• 月之暗面的Kimi K2.5以4.02万亿Token位列第二。
• 中国企业在排名前五的模型中占据三席（MiniMax、月之暗面、DeepSeek），合计贡献了近三分之二的流量份额。
• 分析认为，此次爆发式增长是由于春节前后国内厂商的集中发力，将中美AI研发的技术差距缩小到历史最低点。
• 国内企业级市场AI应用也呈现增长，2025年下半年中国企业日均Token使用量较上半年增长263%。

相关链接：

• OpenRouter ：​中国 AI 模型全球用量首超美国，MiniMax 等国产力量霸榜

Chinese AI Models Surpass US in OpenRouter Usage for the First Time, MiniMax M2.5 Leads Globally

Brief:

• According to the latest data from the online AI hosting platform OpenRouter, Chinese open-source models, represented by MiniMax, Moonshot AI, Zhipu AI, and DeepSeek, collectively surpassed US models in global token usage for the first time in February this year. 🚀
• From February 9th to 15th, Chinese models' weekly call volume reached 4.12 trillion tokens, exceeding US models' 2.94 trillion tokens for the first time.
• In the following week, Chinese models' weekly call volume further increased to 5.16 trillion tokens, marking a staggering 127% growth within three weeks.
• In OpenRouter's February rankings, MiniMax M2.5, launched just two weeks prior, topped the global chart with 4.55 trillion tokens in call volume. 🥇
• Moonshot AI's Kimi K2.5 ranked second with 4.02 trillion tokens.
• Chinese companies secured three spots among the top five models (MiniMax, Moonshot AI, DeepSeek), collectively contributing nearly two-thirds of the total traffic share.
• Analysts attribute this explosive growth to the concentrated efforts of domestic manufacturers around the Chinese New Year, narrowing the technological gap in AI R&D between China and the US to an all-time low.
• AI application in China's enterprise market also showed growth, with daily average token usage by Chinese enterprises increasing by 263% in the second half of 2025 compared to the first half. 📈

Related Links:

• OpenRouter: Chinese AI Models Globally Surpass US in Usage for the First Time, Domestic Powers Like MiniMax Dominate Rankings

快手可灵3.0登顶全球文生视频榜单，彰显中国AI领先实力

简报：

• 快手自研的可灵3.0系列模型（Kling3.0Pro）在2026年2月26日被全球权威AI基准测试机构Artificial Analysis评为文生视频赛道全球第一。
• 可灵3.0Pro以1240分（或1241分）的Arena ELO基准评分位列榜首。
• 在全球前15名的顶尖模型中，可灵家族共有7款模型上榜，形成领先态势。
• 业内分析指出，可灵3.0在视频真实感、画面一致性及可控性上达到行业领先水平。
• 此次登顶标志着AI视频生成技术从“玩具”向“生产力工具”的质变，具备进入影视工业级视觉生产核心环节的能力。

相关链接：

• 可灵3.0登顶文生视频赛道全球第一
• 全球 AI 文生视频赛道“华山论剑”：快手可灵 3.0 系列傲视群雄，前 15 名占 7 席

Kuaishou's Kling 3.0 Tops Global Text-to-Video Ranking, Showcasing China's Leading AI Prowess 🏆

Brief:

• Kuaishou's self-developed Kling 3.0 Pro model was ranked as the world's number one in the text-to-video domain by the globally authoritative AI benchmark testing institution, Artificial Analysis, on February 26, 2026. 🚀
• Kling 3.0 Pro secured the top spot with an Arena ELO benchmark score of 1240 (or 1241) points.
• The Kling family demonstrated a dominant presence, with 7 models making it into the global top 15 leading models.
• Industry analysis highlights that Kling 3.0 has reached industry-leading levels in terms of video realism, visual consistency, and controllability.
• This achievement marks a qualitative transformation for AI video generation technology, evolving from a "toy" into a "productivity tool," and demonstrating its capability to integrate into the core processes of industrial-grade visual production for film and television. 🎬

Related Links:

• Kling 3.0 Tops Global Text-to-Video Ranking
• Global AI Text-to-Video Showdown: Kuaishou's Kling 3.0 Series Dominates, Securing 7 of Top 15 Spots

英伟达年收入飙升至2160亿美元，黄仁勋宣告智能体AI迎来转折点

简报：

• 英伟达2026财年年度收入飙升65%至2160亿美元，第四季度总营收达681.3亿美元，超出华尔街预期。
• 英伟达首席执行官黄仁勋宣布，智能体人工智能（Agentic AI）已达到技术转折点，标志着AI从生成式对话转向能解决实际问题的自主行动，并预测2026年是“个人智能体元年”。
• OpenAI已与OpenClaw创始人Peter Steinberger达成协议，Steinberger将加入OpenAI，致力于将具备自主规划与跨平台执行能力的AI代理推向全球。
• 黄仁勋在财报会议上透露了收购Groq后的核心整合计划，强调Groq将补齐AI推理阶段的短板，实现超低延迟解码能力，并将其重要性与当年收购Mellanox相提并论。
• 英伟达计划在即将到来的GTC大会上披露Rubin芯片细节，并可能进军笔记本芯片市场，同时广发证券预测GTC大会将推出名为“LPX机架”的混合计算方案，集成256个LPU单元。

相关链接：

• 英伟达年收入增至2160亿美元，黄仁勋称智能体AI已达转折点
• 补齐 AI 推理拼图：英伟达黄仁勋揭秘 Groq LPU 整合路线图
• AI熱潮不退燒！輝達Q4表現超乎預期 營收681億美元創新高

NVIDIA's Annual Revenue Soars to $216 Billion, Jensen Huang Declares Agentic AI at a Turning Point

Brief:

• NVIDIA's FY2026 annual revenue surged by 65% to $216 billion, with Q4 total revenue reaching$68.13 billion, exceeding Wall Street expectations. 🚀
• NVIDIA CEO Jensen Huang announced that Agentic AI has reached a technological turning point, marking a shift from generative conversations to autonomous actions capable of solving real-world problems. He predicted 2026 to be the 'Year of the Personal Agent.' 🤖
• OpenAI has reached an agreement with OpenClaw founder Peter Steinberger, who will join OpenAI to focus on bringing AI agents with autonomous planning and cross-platform execution capabilities to the global stage.
• During the earnings call, Huang revealed the core integration plan following the acquisition of Groq, emphasizing that Groq will address shortcomings in the AI inference stage, enabling ultra-low latency decoding capabilities. He likened its importance to the acquisition of Mellanox years ago.
• NVIDIA plans to disclose details of its Rubin chip at the upcoming GTC conference and may enter the laptop chip market. Meanwhile, GF Securities predicts that the GTC conference will unveil a hybrid computing solution called 'LPX Rack,' integrating 256 LPU units. 💡

Related Links:

• NVIDIA's Annual Revenue Increases to $216 Billion, Jensen Huang States Agentic AI Has Reached a Turning Point
• Completing the AI Inference Puzzle: NVIDIA's Jensen Huang Reveals Groq LPU Integration Roadmap
• AI Hype Unabated! NVIDIA's Q4 Performance Exceeds Expectations, Revenue Hits Record $68.1 Billion

美国AI巨头下周白宫签署能源自律承诺

简报：

• 包括OpenAI、亚马逊、微软、谷歌、Meta、xAI及甲骨文在内的多家技术巨头，拟于下周在白宫签署一份正式书面文件。
• 该文件核心承诺企业将自行供应或购买人工智能数据中心所需的电力。
• 此举旨在从基础设施层面解决AI扩张与公共资源分配之间的矛盾，避免将技术研发的成本溢出转嫁至普通消费者。
• 这一承诺标志着AI竞争进入“能源自主”新阶段，巨头们正通过深度布局核能、绿电及微电网技术，试图摆脱对传统公用电网的过度依赖。

相关链接：

• OpenAI、微软等巨头下周拟赴白宫签署承诺:自主解决AI数据中心电力供应

US AI Giants to Sign Energy Self-Regulation Pledge at White House Next Week 🏛️

Briefing:

• Multiple tech giants, including OpenAI, Amazon, Microsoft, Google, Meta, xAI, and Oracle, are expected to sign a formal written document at the White House next week.
• The core commitment of this document is that these companies will either self-supply or purchase the electricity required for their AI data centers. ⚡
• This initiative aims to resolve the conflict between AI expansion and public resource allocation at the infrastructure level, preventing the spillover costs of technological R&D from being passed on to ordinary consumers.
• This pledge marks a new phase of "energy autonomy" in AI competition, as these giants are strategically deploying nuclear power, green electricity, and microgrid technologies in an attempt to reduce their over-reliance on traditional public utility grids. 🤝

Related links:

• OpenAI, Microsoft, and Other Giants to Sign Pledge at White House Next Week: Self-Supply AI Data Center Power

阿里千问进军AI硬件，首款AI眼镜MWC 2026发布

简报：

• 阿里巴巴旗下个人AI助手“千问”宣布进军AI硬件领域，计划于2026年推出AI眼镜、AI指环及AI耳机等产品。
• 千问首款AI眼镜将在2026年世界移动通信大会（MWC）上揭晓，并定于3月2日开启全渠道预约。
• 此举旨在构建“软硬一体”的AI生态，通过硬件传感器捕获物理世界信息，实现AI助手脱离手机终端。
• 千问APP原有的生活服务能力将无缝接入AI眼镜等终端。
• 阿里内部人士确认，即将发布的“千问AI眼镜”与现有“夸克AI眼镜”出自同一团队，未来将统一以“千问（Qwen）”命名，并为已购夸克眼镜用户同步更新功能。

相关链接：

• 阿里千问宣布进军AI硬件领域，首款AI眼镜将亮相2026MWC
• 阿里确认夸克与千问AI眼镜为“同根生”：同一个团队推出的不同系列

Alibaba's Qwen Ventures into AI Hardware, First AI Glasses to Launch at MWC 2026 🚀

Brief:

• Alibaba's personal AI assistant "Qwen" has announced its foray into the AI hardware sector, with plans to launch products such as AI glasses, AI rings, and AI headphones in 2026.
• Qwen's first AI glasses 👓 will be unveiled at the 2026 Mobile World Congress (MWC) and are scheduled to open for pre-orders across all channels on March 2nd.
• This strategic move aims to build a "software-hardware integrated" AI ecosystem, enabling AI assistants to operate independently of mobile phones by using hardware sensors to capture information from the physical world.
• The Qwen APP's existing lifestyle service capabilities will be seamlessly integrated into AI glasses and other devices.
• Alibaba insiders confirmed that the upcoming "Qwen AI Glasses" and the existing "Quark AI Glasses" originate from the same team. In the future, they will be unified under the "Qwen" brand, and functionality updates will be provided simultaneously to existing Quark glasses users. 💡

Related Links:

• Alibaba's Qwen Announces Entry into AI Hardware Sector, First AI Glasses to Debut at MWC 2026
• Alibaba Confirms Quark and Qwen AI Glasses Share "Common Roots": Different Series from the Same Team