AI News AI资讯 4h ago Updated 2h ago 更新于 2小时前 48

Kr Star Evening: DeepSeek Increases Capital by 65% This Year; Musk: FSD is the Core Driver of Tesla's Car Demand 氪星晚报 |DeepSeek今年已增资65%;马斯克:FSD是拉动特斯拉汽车需求的核心驱动力

Alibaba Cloud’s Zhenwu M890 super-node has successfully adapted to the Qwen3.8 model, marking the first domestic super-node to run a large language model with over 2 trillion parameters. The Zhenwu M890 is now live on Alibaba Cloud’s Bailian platform, providing inference services for the massive-scale Qwen3.8 model. This achievement highlights significant progress in domestic hardware-software co-optimization for ultra-large parameter models. 阿里真武M890超节点成功适配Qwen3.8,成为国内首个运行超2万亿参数大模型的超节点,标志着国产算力基础设施在超大模型推理服务上的重大突破。 阅文集团推出“Buddy系列”文创AI智能体,微盟内测“微盟星元”AI原生经营系统,显示AI正从通用工具向垂直行业(内容创作、电商零售)的深度业务流渗透。 飞书aily升级支持自主工作与团队共享智能体,结合马斯克强调FSD为特斯拉核心驱动力,表明Agent自主协作与具身智能应用正在加速落地并重塑商业价值逻辑。

72
Hot 热度
65
Quality 质量
70
Impact 影响力

Analysis 深度分析

TL;DR

  • Alibaba Cloud’s Zhenwu M890 super-node has successfully adapted to the Qwen3.8 model, marking the first domestic super-node to run a large language model with over 2 trillion parameters.
  • The Zhenwu M890 is now live on Alibaba Cloud’s Bailian platform, providing inference services for the massive-scale Qwen3.8 model.
  • This achievement highlights significant progress in domestic hardware-software co-optimization for ultra-large parameter models.

Why It Matters

This development is critical for AI practitioners and researchers as it demonstrates the viability of domestic infrastructure in handling extreme-scale models that were previously difficult or impossible to deploy efficiently. It signals a shift towards localized, high-performance computing solutions capable of supporting next-generation AI applications without reliance on foreign hardware constraints.

Technical Details

  • Hardware Platform: Alibaba Cloud’s Zhenwu M890 super-node, designed for high-throughput and low-latency inference.
  • Model Compatibility: Successfully adapted to Qwen3.8, a large language model exceeding 2 trillion parameters.
  • Deployment Status: The service is operational on the Alibaba Cloud Bailian platform, offering public inference capabilities.
  • Significance: Represents the first successful deployment of a >2T parameter model on a domestic super-node, showcasing advanced system integration and optimization techniques.

Industry Insight

  • Infrastructure Localization: The success of the Zhenwu M890 suggests that domestic chip manufacturers are closing the gap in supporting ultra-large models, reducing dependency on external hardware ecosystems.
  • Scalability Benchmark: This sets a new benchmark for what domestic cloud infrastructure can achieve, encouraging further investment in R&D for large-scale model deployment.
  • Commercial Viability: By making such powerful models accessible via Bailian, Alibaba is lowering the barrier for enterprises to experiment with and integrate state-of-the-art AI capabilities into their workflows.

TL;DR

  • 阿里真武M890超节点成功适配Qwen3.8,成为国内首个运行超2万亿参数大模型的超节点,标志着国产算力基础设施在超大模型推理服务上的重大突破。
  • 阅文集团推出“Buddy系列”文创AI智能体,微盟内测“微盟星元”AI原生经营系统,显示AI正从通用工具向垂直行业(内容创作、电商零售)的深度业务流渗透。
  • 飞书aily升级支持自主工作与团队共享智能体,结合马斯克强调FSD为特斯拉核心驱动力,表明Agent自主协作与具身智能应用正在加速落地并重塑商业价值逻辑。

为什么值得看

本文揭示了AI技术从底层算力突破到上层垂直应用落地的完整链条,特别是国产超节点对万亿参数模型的支持能力,以及AI Agent在办公、内容、零售等具体场景中的产品化进展。对于从业者而言,理解这些技术如何转化为具体的商业驱动力(如特斯拉FSD、微盟对话即经营)至关重要。

技术解析

  • 超节点与大模型适配:阿里云真武M890超节点成功适配Qwen3.8(超2万亿参数),并上线百炼平台提供推理服务。这是国内首个实现此类规模运行的超节点,证明了国产硬件在支撑超大语言模型高效推理方面的技术成熟度。
  • AI原生交互范式转变:微盟“微盟星元”将传统GUI界面转型为LUI(语言用户界面),通过自然语言对话驱动经营。其核心技术在于将零售行业的Know-how拆解为标准化的Skill模块,使Agent具备任务拆解和Skill调度能力,实现“对话即经营”。
  • 智能体自主协作机制:飞书aily升级后,能够在授权范围内基于多模态信息(消息、文档、日历等)主动跟进任务,并与团队成员或其他智能体分工协作。这体现了Agent从单点辅助向多智能体协同工作的演进。
  • 垂直领域AI工具矩阵:阅文发布“Buddy系列”,涵盖作家助手、漫剧助手和版权助手,形成覆盖内容创作、IP改编、版权开发的一体化AI工具矩阵,展示了AI在特定创意产业工作流中的深度整合能力。

行业启示

  • 算力国产化进入深水区:阿里真武超节点支持万亿参数模型运行,意味着中国科技巨头在高端AI算力基础设施上已具备独立支撑前沿大模型应用的能力,减少了对海外算力的依赖,为国内大模型生态提供了坚实底座。
  • AI应用向“业务内核”下沉:无论是微盟的“对话即经营”还是阅文的“一体化AI工具矩阵”,AI不再仅仅是效率提升工具,而是开始重构业务流程和交互方式。企业应关注如何将行业Know-how转化为AI可调用的Skill,以构建竞争壁垒。
  • 具身智能与自动驾驶的商业闭环验证:马斯克关于FSD拉动需求的观点,结合钧舵机器人、乐享智能等获投新闻,表明具身智能和自动驾驶正从技术演示走向明确的商业变现路径。投资者和企业应重点关注那些能清晰界定AI价值主张并产生直接收入增长的细分赛道。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

Open Source 开源 LLM 大模型 Funding 融资 Autonomous Driving 自动驾驶