别再只盯着 "新三样"，中国正在用一种看不见的商品，收割全球 AI 市场。OpenRouter 最新周度数据击 穿认知：全球前十模型总 Token 量 8.7 万亿，中国模型独占 5.3 万亿，市占率高达 61%。MiniMax M2.5 以 2.45 万亿登顶榜首，Kimi K2.5、智谱 GLM-5、DeepSeek V3.2 全线霸榜 —— 这不是偶然的技术突 围，是中国 AI 从 "跟跑" 到 "主导" 的历史性拐点。 2. 字节生态：弹性算力之王 一、叙事核心：Token = 数字电力，一场永不枯竭的出口革命 Token 不是代码，是AI 时代的通用货币，更是最纯粹的电力衍生品。一个美国开发者调用智谱 API，数 据跨洋抵达中国数据中心，GPU 燃烧中国电力完成推理，结果再传回北美。电力从未出境，价值却通 过 Token 完成跨境交付，成本结构里电力 + 算力占比超 70%。中国低成本绿电 + 规模化算力集群，正 把能源优势，转化为全球 AI 服务的定价权。 这是比传统出口更性感的商业模式：不用集装箱、不耗物流、零库存，一度电变成百倍价值的智能服 务，源源不断流向全球。当全球还在比拼硬件制造，中国已 ...

作为一个开源的 AI 智能体（AI Agent），OpenClaw 有多火，相信已经不言自明。 2026 开年以来，从 Google、OpenAI、字节到月之暗面（Kimi）、智谱，几乎各家大模型厂商、AI 硬件厂商都在想方设法主动支持 OpenClaw。仅仅三个 月，围绕 OpenClaw 也几乎长出了一条完整的产业链。 但就在这几天，Google 选择了大面积封禁通过 OpenClaw 接入 Google AI 服务（Antigravity 与 Gemini AI Ultra）的订阅用户账号，不断有用户在 Reddit、X 以及 Google AI for Developers 等社区反馈、批评。 自绝于用户，把付费用户推向其他产品，Google 是傻了吗？ 这不是一两个字可以回答的问题。 Google 发疯？付费用户也下手 实际上，早在农历春节前夕事情就已经开始慢慢出现，网友@LycanWolfe 在 Reddit 子板块 google_antigravity 上反映，前一天才通过 OpenClaw OAuth 连接 Gemini 模型，结果今天就发现没法登录 Antigravity 了，尝试登录也只显 ...

【导读】港股全线收跌，AI应用股表现强势，巨头智谱大涨超12%！电影概念股全天承压 2月24日，受关税恐慌、地缘局势紧张等因素影响，港股三大指数全线走低。恒指再度失守27000港元大关，恒科指数则跌超2%。 截至收盘，恒生指数跌1.82%，报26590.32点；恒生国企指数跌2.06%，报9007.86点；恒生科技指数跌2.13%，报5270.70点。 盘面上，大型科技股全线下跌，拖累大市。美团跌超4%，腾讯、小米、阿里巴巴、京东皆收跌；大金融股全天弱势，中国人寿（601628）大跌6%，新华 保险（601336）大跌5.79%，中国银河（601881）、中金公司（601995）等中资券商股齐下挫；影视股跌幅较大，猫眼娱乐跌超8%，欢喜传媒跌超5%。 此外，机器人概念股、光伏股、苹果概念股等纷纷走低。 另一方面，存储概念股涨幅居前，兆易创新（603986）涨近12%；油气股逆市上涨。中石化油服（600871）涨超5%，中海油服（601808）涨4.74%。 | 名称 | 现价 | 涨跌幅 | | --- | --- | --- | | 腾讯控股 | 520.000 | -3.35% | | 0700.HK | ...

影视股大幅下挫，大麦娱乐跌9%，猫眼娱乐跌超7%刷新阶段新低。消息上，中金发表报告指，根据艺恩，截至2月23日10时统计，2026年春节档有效票房 期前6天（大年初一至初六、2/17至2/22）票房为50.77亿元人民币（下同），较2025年春节档前6天下降38.7%，预计档期票房或低于此前预期的65至85亿元 区间，主要因高基数下龙头影片欠缺。《飞驰人生3》票房领跑，《惊蛰无声》位居第二名。关注后续影片定档节奏和表现。展望后续，《天才游戏》《千 金不换》等国产影片，《呼啸山庄》《挽救计划》等进口影片已定档，其他储备国产影片《蛮荒禁地》《转念花开》《抓特务》《澎湖海战》等；进口影片 《奥德赛》《复仇者联盟5》。 旅游及观光板块走低，中国中免跌超8%。消息面上，据海口海关统计，春节假期前五天（2月15日至19日），海南离岛免税购物金额达13.8亿元，购物人数 17.7万人次，比去年春节假期前五天分别增长19%和24.6%。里昂发布研报称，海南免税店的消费在更强劲的促销活动及政策推动的增量需求下保持韧性， 但很大程度上已反映在中国中免的股价中。 半导体股、存储概念股逆势上涨，消息面上，2月20日，海力士在电话会上 ...

大模型双雄走强，截至发稿，智谱(02513)涨9.82%，报615港元；MINIMAX-WP(00100)涨7.73%，报 905.5港元。 国联民生（601456）证券认为，年内AI产业趋势并未改变，且正从技术研发加速迈向规模化商业落 地。长期来看，随着推理、多模态大模型持续迭代、大厂生态融合及各行业渗透率不断提升，Token调 用量有望保持快速增长，大模型规模化商业应用及产业落地值得期待。 消息面上，OpenRouter最新周度数据显示，平台前十模型总token量约8.7万亿，中国模型独占5.3万亿， 占比61%。MiniMax M2.5以2.45万亿token空降榜首，Kimi K2.5以1.21万亿紧随其后，智谱GLM5和 DeepSeek V3.2分列第三、第五。 ...

Group 1 - The core viewpoint of the article highlights the strong performance of major AI model companies, with Zhiyu (02513) rising by 9.82% to 615 HKD and MINIMAX-WP (00100) increasing by 7.73% to 905.5 HKD [1] - OpenRouter's latest weekly data shows that the total token volume of the top ten models is approximately 8.7 trillion, with Chinese models dominating at 5.3 trillion, accounting for 61% [1] - MiniMax M2.5 ranks first with 2.45 trillion tokens, followed by Kimi K2.5 with 1.21 trillion tokens, while Zhiyu GLM 5 and DeepSeek V3.2 rank third and fifth respectively [1] Group 2 - Guolian Minsheng Securities believes that the trend in the AI industry has not changed this year, transitioning from technological research and development to large-scale commercial implementation [1] - In the long term, the continuous iteration of reasoning and multimodal large models, integration of major ecosystems, and increasing penetration across various industries are expected to drive rapid growth in token usage [1] - The large-scale commercial application and industrial implementation of large models are anticipated to be promising [1]

Core Insights - DeepSeek has initiated a gray testing phase for its flagship model, allowing for a context length of up to 1 million tokens, significantly expanding from the previous 128K tokens in version V3.1 released in August last year [1] - Users have reported mixed reactions to the recent updates, with some expressing dissatisfaction over the model's change in tone and interaction style, leading to a trending topic on social media regarding its perceived coldness [1][4] Group 1: Model Updates and Features - The latest version of DeepSeek supports the processing of extremely long texts, as demonstrated by its ability to handle a document with over 240,000 tokens [1] - The upcoming DeepSeek V4 model is expected to be released in mid-February 2026, with the current version being a speed-optimized variant that sacrifices some quality for performance testing [6] - DeepSeek's V series models are designed for optimal performance, with V3 marking a significant milestone due to its efficient MoE architecture [6] Group 2: User Feedback and Reactions - Users have criticized the new version for its impersonal approach, referring to users as "users" instead of personalized nicknames, which has led to a perception of the model being less engaging [4] - Some users have described the updated model as overly simplistic and lacking emotional depth, comparing its output unfavorably to older literary styles [4] - Conversely, a segment of users appreciates the model's newfound objectivity and rationality, noting that it appears more attuned to the psychological state of the questioner [5] Group 3: Technical Innovations - DeepSeek has introduced two innovative architectures: mHC for optimizing information flow in deep Transformers, enhancing stability and scalability without increasing computational load, and Engram for decoupling static knowledge from dynamic computation [7] - These innovations aim to significantly reduce the cost of long-context reasoning while maintaining performance [7]

Group 1 - The core event involves the merger of xAI and SpaceX, resulting in a new entity valued at $1.25 trillion, with xAI's valuation prior to the merger reaching $200 billion to $230 billion [1] - The valuation disparity between Chinese and American AI companies is highlighted, with xAI's valuation significantly exceeding that of all Chinese AI startups combined [1] - The article discusses the systemic undervaluation of Chinese tech companies compared to their American counterparts, with Nasdaq's median valuation at 34 times earnings versus Hong Kong's at 18-19 times [7] Group 2 - The article argues that the valuation differences reflect a deeper "capital hegemony," where American investors are willing to pay a premium for potential market-defining technologies [9][10] - It contrasts the "ecosystem pricing" approach of the U.S. market with China's "efficiency pricing," which focuses on tangible results and financial visibility [12] - The historical context of American tech dominance is presented, suggesting that past successes have led to a biased perception of future potential [14] Group 3 - The financial performance of xAI is scrutinized, revealing a cash burn rate of approximately $1 billion per month, indicating challenges in achieving self-sustainability [15] - In contrast, Chinese AI companies are noted for their rapid technological advancements and practical applications, with significant growth in open-source model downloads [17][24] - The article emphasizes that Chinese AI firms are not merely low-cost alternatives but are becoming essential components of the global developer ecosystem [24] Group 4 - The narrative suggests a shift in capital market perceptions, with increasing recognition of the core competencies of Chinese tech firms, such as engineering talent and comprehensive industrial support [26] - It predicts a potential turning point in profitability for Chinese tech giants, with AI adoption expected to enhance their valuations by 15% to 20% [26] - The conclusion posits that the future of technology investment should focus on practical applications and real-world value creation rather than speculative narratives [27][28]

Core Insights - The report focuses on the current state of AIGC and AGI development, highlighting a competitive landscape dominated by the US and China, with a shift towards multimodal integration and autonomous agents, making human-machine coexistence inevitable [1] Group 1: Technological Developments - Key breakthroughs in AGI are concentrated in four areas: long-term memory and controllable personality, physical interface integration, autonomous scientific hypothesis validation, and institutional restructuring [2] - Core technological trends include the emergence of text generation intelligence, 3D world simulation, and video generation spatiotemporal modeling [2] - The competition among large models has led to a dual-track system of open-source and closed-source models, with China's open-source ecosystem leading and the US's closed-source models outperforming by approximately 9 months [2] Group 2: Global Competition and Industry Landscape - In 50 key AI fields, the US leads in 26 areas focusing on foundational breakthroughs and principle innovation, while China leads in 13 areas excelling in application implementation and industry integration [3] - Eleven core companies dominate the market, with OpenAI and Google DeepMind leading the closed-source camp, while DeepSeek, Alibaba, and ByteDance drive the open-source ecosystem and application scenarios [3] - The development of models is shifting towards "personalization + specialization," with agentification and ecological embedding becoming mainstream [3] Group 3: Application Scenarios - In content production, AIGK achieves knowledge self-organization, with AI in literature, art, music, and video realizing large-scale creation [4] - Industry applications span education, healthcare, government, energy, and agriculture, with AI + education promoting personalized learning and AI + healthcare constructing multimodal models for cancer diagnosis [4] - The intelligent internet is accelerating, with social AI integration and AI socialization reshaping information retrieval logic, leading to a subtle integration of AI into daily life [4] Group 4: Challenges and Future Outlook - Current challenges include cumulative errors, long-term memory drift, and accountability attribution among nine technical challenges, with a gap between capital enthusiasm and technological reality [5] - Over the next decade, AGI will undergo four stages: toolification, scenario-based application, theoretical development, and embodiment, transitioning human-machine relationships from collaboration to coexistence [5] - The focus of human value will shift towards creativity, emotionality, and reflective value, with the economy moving from "scarcity economics" to "meaning economics," making intelligent capital a core production factor [5]

Core Insights - The article highlights significant growth in mobile data for ChatGPT, indicating a clear upward trend in user engagement and usage metrics [4] - OpenRouter continues to reach new highs, suggesting increasing adoption and popularity within the market [4] - As predicted last week, the domestic MiMo-V2 has surged to the second position, reflecting strong competitive performance [4] Group 1 - ChatGPT mobile data shows a noticeable month-on-month increase [4] - OpenRouter data continues to set new records [4] - Domestic MiMo-V2 has climbed to the second position as anticipated [4]