智能体方面，2026年初，名为OpenClaw的开源智能体工具在开发者社区和各大技术论坛引起较大轰 动，它不再是过去那个只会陪聊的聊天机器人，而是进化成了能够接管电脑、协助人类完成具体办公任 务的"办公搭子"，初步展示了生产力大变革的形象。多模态方面，字节发布的SeeDance2.0让大模型具 有了影视行业的生产力，从此前"生成一段画面"走向"完成一个作品"，生成15秒视频的可用率从此前 20%提升至90%，提升效率降低成本，有望推动漫剧等行业进入规模化发展阶段。 中银国际认为，随着大模型能力提升，这两条演进路径逐步清晰：Agent通过理解用户意图、拆解复杂 任务，搭载MCP和Skills等工具，快速覆盖商业办公、法律和金融领域应用；多模态有望通过生成仿真 数据，进一步用于加速具身智能等场景。 算力涨价成为新迹象，凸显算力供应瓶颈，算力产业链有望持续受益 中银国际发布研报称，2026年春节前后，国内外主要AI模型纷纷完成重大升级。随着大模型能力提 升，智能体与多模态两条演进路径逐步清晰。与此同时，智谱(02513)节前发布GLM Coding Plan价格调 整函，这一涨价预示着优秀的大模型企业不缺需求，有望 ...

截至2026年2月25日 09:45，中证人工智能产业指数(931071)下跌0.05%。成分股方面涨跌互现，星宸科 技领涨2.16%，均胜电子上涨2.09%，慧博云通上涨2.08%；芯原股份领跌，澜起科技、星环科技跟跌。 人工智能ETF(515980)盘中成交4378.14万元。拉长时间看，截至2026年2月24日，人工智能ETF近一年累 计上涨56.44%。 中信证券研报称，2026春节期间，国产大模型token调用量井喷，截至2月22日此前一周的全球大模型 token用量中，前三甲均为国产大模型。中信证券认为，token的爆发式增长，本质上反映出AI推理需求 的指数级扩容，而国产算力凭借着成本优势及不断完善的生态，有望在基础设施层逐步占据主导。 华富人工智能ETF（515980）跟踪中证人工智能产业指数，依托标的指数三层闭环编制逻辑+季度调仓 紧跟产业趋势的优势，是布局A股AI的优质小宽基标的。在同类AI相关指数中具备较高的AI应用含量， 将40%的权重聚焦于应用端，同时也保持了60%的算力基础设施这一产业共识核心。结合春季行情日历 效应，当前市场震荡回调中，投资者可密切关注人工智能ETF，把握AI全产 ...

截至2026年2月25日 09:32，中证人工智能主题指数(930713)下跌0.30%。成分股方面涨跌互现，星宸科 技领涨4.17%，协创数据上涨1.23%，拓维信息上涨1.21%；芯原股份领跌4.09%，昆仑万维（维权）下 跌2.75%，北京君正下跌2.47%。AI人工智能ETF(512930)下跌0.40%，最新报价2.27元。 资金流入方面，AI人工智能ETF最新资金净流入3194.73万元。拉长时间看，近4个交易日内有3日资金 净流入，合计"吸金"1.25亿元，日均净流入达3118.97万元。（数据来源：Wind） 上个马年沪指涨近60%！新年新福利来了，炒股排面要拉满，新号启幸运>> 消息面上，全球最大AI开源社区Hugging Face公布新一期开源大模型榜单，阿里最新开源的原生多模态 模型千问3.5登顶全球榜首。在最新一期Hugging Face模型榜单前十中，有8款来自中国团队，涵盖阿里 巴巴、智谱、MiniMax、月之暗面等多家机构。 中信证券研报指出，2026春节期间，国产大模型token调用量井喷，截至2月22日此前一周的全球大模型 token用量中，前三甲均为国产大模型。token的 ...

昨日（2月24日），港股科技方向走低。国证港股通科技指数(987008.SZ)下跌2.47%，该指数成分股 中，华虹半导体上涨超3%，小鹏汽车-W上涨超1%，百济神州上涨0.65%，瑞声科技上涨0.31%，同程 旅行下跌0.09%。 相关ETF方面，Wind数据显示，港股科技ETF天弘（159128）昨日全天成交额为1.06亿元，溢折率为 0.69%，盘中获净申购超8000万份。 资金流向方面，Wind数据显示，2月13日该ETF净流入额为6655.05万元，截至2月13日，已连续25个交 易日获资金净流入，累计净流入额为8.49亿元。该ETF最新流通份额为27.28亿份，最新流通规模为 24.32亿元。 港股科技ETF天弘（159128）紧密跟踪港股通科技指数，成分股均属于沪深港通标的，不受QDII额度 的限制，可T+0交易，该ETF还配备了场外基金联接A（024885）、联接C（024886）。 消息面上，据智通财经，全球最大AI开源社区Hugging Face公布新一期开源大模型榜单，阿里巴巴最新 开源的原生多模态模型千问3.5登顶全球榜首。在最新一期Hugging Face模型榜单前十中，有8款来自中 ...

| Models 2,647,509 Filter by name | Full-text search * Inference Available 1J Sort: Trending | | --- | --- | | Qwen/Qwen3.5-397B-A17B | Z zai-org/GLM-5 | | [% Image-Text-to-Text · .: 403B · Updated about 16 hou ... · ¿ 303k · + · ♡ 952 | To Text Generation · .: 754B · Updated 11 days ago · · 2 179k · 4 · ♡ 1.48k | | 2 nvidia/personaplex-7b-v1 | 글 Nanbeige/Nanbeige4.1-3B | | 11 Audio-to-Audio · Updated 8 days ago · 2 539k · ♡ 2.18k | To Text Generation · .: 4B · Updated 2 days ago · ¿ 178k · · ♡ 756 | | @ Mi ...

可选消费板块下跌，中国中免（601888）跌超10%。 2月24日，港股收盘走低，恒生指数跌1.82%，恒生科技指数跌2.13%。 | 香港股票 △ | | | | --- | --- | --- | | 恒生指数 | 恒生科技 | 恒生生物科技 | | 26590.32 | 5270.70 15654.19 | | | -491.59 -1.82% -114.65 -2.13% -541.42 -3.34% | | | | 港股通消费 | 港股红利 | 港股现金流 | | 1704.19 | 2318.72 | 2555.21 | | -22.73 -1.32% -2.80 -0.12% -41.84 -1.61% | | | 互联网科技股集体走低，阿里巴巴跌2.7%，金山云跌5%，快手、小米等跌逾2%。 | 代码 | 名称 | 现价 | 涨跌 | 涨跌幅 ▲ | 成交额 | | --- | --- | --- | --- | --- | --- | | 0268 | 全蝶国际 | 9.970 c | -0.840 | -7.77% | 7.21亿 | | 0020 | 商沥-W | 2.610 c | - ...

Group 1 - The company has secured orders for its connectivity business until the fourth quarter of 2026, indicating strong demand and a robust order backlog [1] - The AI high-speed optical module production line is operating at full capacity 24 hours a day, reflecting the company's commitment to meeting market needs [1] Group 2 - Alibaba Cloud's Coding Plan supports various models including Q&A 3.5, GLM-4.7, and Kimi-K2.5, showcasing advancements in AI technology [1] - Zhiyuan released a technical report on GLM-5, providing full disclosure of technical details, which may enhance transparency and trust in their AI solutions [1]

Group 1 - The core viewpoint is that China's technology industry has initiated an AI "hundred model war," with a series of domestic AI models being released and upgraded, indicating that the impact of Chinese technology on the world is just beginning [1][17]. - Multiple Chinese tech companies, including Moonlight Dark Side, Zhiyu, and Xiyu Technology, have launched new generation AI large models since before the Spring Festival, with ByteDance releasing the Doubao Model 2.0 series on February 14, enhancing its ability to perform complex real-world tasks [3][5]. - The Seedance 2.0 video generation model was officially launched by ByteDance, showcasing its capabilities during the 2026 Spring Festival Gala, where it animated a traditional Chinese ink painting, demonstrating advanced motion and camera control [7][9]. Group 2 - The release of Seedance 2.0 has sparked global discussions, with industry insiders noting that the focus on video and image generation models represents the next stage of AI technology, with breakthroughs occurring sooner than previously expected [11]. - The recent upgrades in domestic AI models emphasize the development of "AI agents," which are designed to perform more complex tasks beyond basic language communication, resembling intelligent assistants [13]. - Alibaba released the new base model Qianwen 3.5 on February 16, which supports mixed visual and text pre-training and significantly enhances capabilities in problem-solving, spatial reasoning, and video understanding [15]. Group 3 - Following the release of domestic models, several Chinese chip companies announced compatibility with these models, indicating that they can run effectively on their chips, achieving expected performance levels [15][17]. - Huawei's Ascend chips are providing training support for domestic large models, showcasing China's rapid progress in AI despite export controls, with innovations emerging from demand-driven needs [17].

Core Insights - The launch of ByteDance's video generation model Seedance 2.0 has gained global popularity among creators due to its multimodal reference system and native audio-video synchronization capabilities [3] - The AI large model industry is experiencing a new wave of innovation, with numerous new models being released both domestically and internationally, including Alibaba's Qianwen 3.5 and OpenAI's GPT-5.3-Codex [3][4] - The rapid development of AI requires a comprehensive innovation ecosystem that integrates various sectors, including finance, healthcare, manufacturing, and transportation, to realize its full potential [4] Industry Developments - The AI large model sector is witnessing a surge in new applications and models, characterized by increasing differentiation and practicality, with many models breaking out of traditional boundaries [4] - Shanghai has strategically positioned itself in the AI landscape by fostering an environment conducive to innovation, including support for startups and a focus on real-world applications [5][6] - The city has achieved a record of five AI companies going public in one month, indicating its competitive edge in the AI sector [5] Innovation Ecosystem - The development of any innovative industry in Shanghai necessitates a systematic and interconnected framework to stimulate vitality and ensure sustainable growth [6] - The concept of "AI for science" and "AI for maths" highlights the interdependence between AI advancements and progress in other scientific fields, further emphasizing the need for a collaborative ecosystem [4]

Core Viewpoint - The release of Qwen 3.5 has sparked significant interest in the global AI industry, showcasing advancements in model efficiency and cost-effectiveness [1] Group 1: Model Performance and Features - Qwen 3.5 has a total of 397 billion parameters, with only 17 billion activated, outperforming the previous trillion-parameter model Qwen 3-Max [1] - The deployment memory usage has decreased by 60%, while the maximum inference throughput has increased by 19 times [1] - The innovative underlying model architecture allows Qwen 3-Plus to offer competitive performance at a low cost of 0.8 yuan per million tokens on Alibaba Cloud's Bailian API, achieving high performance comparable to Gemini 3 at less than 5% of the price [1] Group 2: Industry Adoption and Compatibility - Major international hardware manufacturers such as NVIDIA, AMD, and Apple have adapted their development frameworks and chips for Qwen 3.5 [1] - Domestic GPU and platform providers, including Huawei Ascend, Moore Threads, and others, have also announced Day 0 compatibility with the new model [1] - Various platforms, both domestic and international, including National Supercomputing Internet Platform, Shuguang Cloud, and others, have quickly integrated the Qwen 3.5 model, offering API services or experience interfaces [1]