衡宇 鱼羊 发自 凹非寺 量子位 | 公众号 QbitAI 3D生成版DeepSeek再上新高度！ 国产、易用、性能强且开源—— 新模型一露面就刷新SOTA，并且 第一时间加入开源全家桶 。 顺时针转个圈圈给大家看，效果是这样： 加上"皮肤"是这样： 再来一个，效果是这样： 肉眼可见，这次妥妥升级变成了更细节的细节控～ 以上效果，都来自 3D大模型明星初创公司VAST ，其刚刚上新的两个基础模型，TripoSG和TripoSF，为团队的最新研发成果。该团队去年3 月开源了TripoSR，在开源3D生成基础模型中爆火全球。 TripoSG ，发布即开源，一露面就刷新开源3D生成模型SOTA，让广大开发者第一时间享受技术进步的成果。 TripoSF ，目前为开源第一阶段，已经用实力证明了自己：横扫一切开源和闭源的现有方法，拿下新SOTA。 你就说秀不秀吧 （手动狗头） ？！ ——但基础模型还只是VAST最近大秀一波技术肌肉的上半程表演。 量子位获悉， 接下来VAST要连续开源一个月，每周都有新开源项目公布 。而TripoSG和TripoSF是开源月里第二周的项目。 在整个开源月里，除了第一波单张图像端到端生成三维 ...

点击 上方 硬AI 关注我们 巴克莱指出，2025年AI行业有足够的算力来支持15亿到220亿个AI Agent。AI行业需从"无意义基准测试"转向实用的Agent产品部署，低推理成本是盈利关键，开源模型将降低 成本。尽管算力看似充足，但高效、低成本Agent产品的专用算力仍有缺口。 硬·AI 作者 |鲍亦龙 编辑 | 硬 AI 继TD Cowen后，巴克莱似乎也开始唱空AI算力。 3月26日，巴克莱发布最新研究称，2025年全球AI算力可支持15-220亿个AI Agent，这足以满足美国和欧盟1亿多白领工作者和超过10亿企业软件许可证的 需求。而同日 TD Cowen分析师称支撑人工智能运算的计算机集群供过于求 。 巴克莱认为现有的AI算力已经足够支持大规模AI代理的部署，主要基于以下三点： 行业推理容量基础 ：2025年全球约有1570万个AI加速器（GPU/TPU/ASIC等）在线，其中40%（约630万个）将用于推理， 而这些推理算力中约一半（310万个）将专门用于 Agent/聊天机器人服务 ； 可支持大量用户 ：根据不同模型的计算需求，现有算力可支持15亿到220亿个AI代理，这足以满足美国和欧 ...

Core Viewpoint - The future of AI will lead every company to become an AI company, blurring the lines between AI and non-AI companies, as AI will transform all aspects of product development and problem-solving [2][10]. Group 1: Company Background and Development - Shen Zhenyu, the founder of Tensor.Art, has a background in AI and has witnessed the evolution of AI algorithms from classic machine learning to modern deep learning techniques [3]. - The company, originally known as QianDao, has transitioned into the AI space with Tensor.Art, which serves as a community and infrastructure for AI model sharing and training [11]. Group 2: Tensor.Art's Positioning and Strategy - Tensor.Art is positioned as a leading platform for AIGC model hosting and sharing, with over 2 million users and more than 500,000 models, generating over 2 million images daily [9]. - The platform aims to create a dual moat through model scale and creator scale, emphasizing that a larger number of models and creators will enhance commercial efficiency [19][20]. Group 3: AI Technology and Market Trends - AI technology is expected to become as fundamental as electricity, necessitating a shift towards numerous fine-tuned models to address specific scenarios rather than relying solely on large models [2][12]. - The company believes that open-source models will dominate the future, as they allow for greater participation from global talent and provide more flexibility for businesses compared to closed-source models [12][16]. Group 4: Competitive Advantages - Tensor.Art's competitive edge lies in its strong hosting capabilities, offering superior inference performance and cost-effectiveness compared to competitors like Civitai [17]. - The platform is designed to support creators in monetizing their models, with revenue-sharing mechanisms similar to those used by popular content platforms [18]. Group 5: Future Directions and Innovations - The company is exploring the integration of video and 3D models into its offerings, recognizing the growing demand for video content generation and the potential for significant market expansion [22][23]. - Tensor.Art is committed to remaining a facilitator of open-source models rather than developing proprietary models, focusing on supporting the broader open-source ecosystem [16].

Core Viewpoint - DeepSeek has released the latest update of its V3 model, named V3-0324, which optimizes performance, user experience, and practicality while maintaining the original technical framework [1]. Group 1: Model Performance - The V3-0324 model has 685 billion parameters, a slight increase from the previous version's 671 billion [1]. - User tests indicate improved performance in generating complex code, solving mathematical problems, and front-end design tasks, with notable enhancements in front-end coding capabilities [2]. - Users have compared the performance improvement of V3-0324 to the upgrade from Sonnet 3.5 to Sonnet 3.6, highlighting its ability to create sophisticated websites with minimal input [2]. Group 2: User Interaction - The new model has disabled the "deep thinking" mode by default, resulting in faster response times suitable for rapid iteration tasks [2]. - The model's natural language processing capabilities have improved, with better context understanding and more human-like responses, reducing mechanical replies [3]. Group 3: Open Source Licensing - V3-0324 continues DeepSeek's open-source tradition, now under the more permissive MIT license, allowing researchers and developers to freely download, modify, and deploy the model [3]. - The updated licensing conditions are expected to attract global developers' attention, despite this upgrade not being the anticipated V4 or R2 version [3]. Group 4: Market Expectations - Analysts suggest that the release timing and features of V3-0324 may indicate it will serve as the foundational model for the upcoming DeepSeek-R2 [3]. - There are market speculations about the early release of DeepSeek-R2, although the official details and release date remain unconfirmed, with expectations set for May [3].

Core Insights - The introduction of DeepSeek-R1 marks a pivotal moment in the AI landscape, bridging the gap between open-source and proprietary models, with significant implications for AI infrastructure and generative AI economics [1][2][8] Open Source vs. Proprietary Models - DeepSeek-R1 has significantly narrowed the performance gap with proprietary models like OpenAI, achieving parity in key reasoning benchmarks despite being smaller in scale [2] - The emergence of DeepSeek is seen as a watershed moment for open-source AI, with models like Llama, Qwen, and Mistral expected to catch up quickly [2][3] - The competitive landscape is shifting, with a vibrant and competitive LLM market anticipated, driven by the open-source model's advancements [2][3] AI Infrastructure and Developer Utilization - DeepSeek-R1 utilizes reinforcement learning (RL) to enhance reasoning capabilities, marking the first successful large-scale implementation of this approach in an open-source model [3][4] - The model's success is expected to democratize access to high-performance AI, allowing enterprises to customize solutions based on their specific needs [3][4] - The shift in AI infrastructure is characterized by a move away from closed models, enabling more control and flexibility for developers [4] New Applications: Large-Scale AI Reasoning - Enhanced reasoning capabilities of DeepSeek open up new application possibilities, including autonomous AI agents and specialized planning systems across various industries [5][6] - The demand for GPU computing is expected to increase due to the accelerated adoption of agent applications driven by DeepSeek [6] - Companies in highly regulated industries will benefit from the ability to experiment and innovate while maintaining control over data usage [6] Generative AI Economics: Changing Cost Dynamics - DeepSeek is driving a trend towards lower costs and higher efficiency in reasoning and training, fundamentally altering the economics of generative AI deployment [7][8] - Models like R1 can be up to seven times cheaper than using proprietary APIs, unlocking previously unfeasible use cases for many enterprises [7] - The economic advantages of open-source models are expected to lead to a broader adoption of AI technologies across various sectors [7][8] Conclusion - DeepSeek represents a significant milestone in the AI industry, enabling open-source models to compete effectively with proprietary alternatives, while emphasizing the importance of high-quality, domain-specific data and labeling for future advancements [8]

以下文章来源于财经杂志 ，作者《财经》杂志 财经杂志 . 《财经》杂志官方微信。《财经》杂志由中国证券市场研究中心主办，1998年创刊，秉承"独立、独家、独到"的新闻理念，以权威性、公正性、专业性报 道见长，是政经学界决策者、研究者、管理者的必读刊物。 打破了几个"刻板共识"。 作者 | 《财经》杂志 来源 | 财经杂志 （ID： i-caijing） 2025年春节的前一周，节日气氛正浓，来自中国杭州的一家初创公司开始让华尔街投资人寝食难安。 杭州量化私募机构幻方旗下的大模型公司DeepSeek（深度求索）发布的一款开源AI模型，在多项测试中表现优于OpenAI的产品，且 研发成本不到600万美元。更让华尔街投资人震惊的是，DeepSeek1月20日发布R1模型，上线六天后同时登顶苹果App Store和谷 歌Play Store全球下载榜首，上线18天内，累计下载量已突破1600万次。随后在2月1日突破3000万大关，成为史上最快达成这一 里程碑的应用。 华尔街的担忧在于，目前投入数十亿美元用于构建大型AI模型的做法可能会打水漂，更廉价的替代方案将让华尔街人工智能的泡沫破 裂。泡沫破裂带来的资本市场危机短 ...