单词学习：![[946f7786-6920-4696-8cbf-581760385173.mp3]]

新闻播报：![[fa7b1aee-4423-43f7-9ca6-fcd50fea60a9.mp3]]

百度发布文心5.0大模型，参数规模超2.4万亿

简报：在2025百度世界大会上，百度创始人李彦宏正式发布了参数规模超2.4万亿的新一代原生全模态大模型——文心大模型5.0，该模型采用超稀疏混合专家架构，在全模态理解、创意写作等方面均有提升，目前已上线百度千帆平台。

相关链接：

• 百度李彦宏发布文心大模型 5.0：参数规模超 2.4 万亿，原生全模态

Baidu Unveils Wenxin 5.0 Large Model with Over 2.4 Trillion Parameters ✨

Brief: At the 2025 Baidu World Conference, Baidu founder Robin Li officially unveiled Wenxin Large Model 5.0, a new generation native full-modal large model with over 2.4 trillion parameters. The model employs an ultra-sparse mixture-of-experts (MoE) architecture, showing significant improvements in full-modal understanding, creative writing, and more. It is currently available on Baidu Qianfan Platform. 🚀 This release promises exciting advancements in AI capabilities. 🧠

Related Links:

• Baidu's Robin Li Releases Wenxin Large Model 5.0: Over 2.4 Trillion Parameters, Native Full-Modal

特斯拉、沃尔沃、奔驰AI大模型率先通过国家备案

简报：

• 特斯拉的xBot客户服务、沃尔沃的小沃智能助手以及奔驰的虚拟助手已通过国家备案或审批，成为全国首批获批上线的外企大模型产品，将为用户提供产品咨询、用车问答等智能服务。

相关链接：

• 全国首批、上海两款外企 AI 大模型通过国家备案，分别来自特斯拉和沃尔沃

Tesla, Volvo, Mercedes-Benz AI Large Models First to Pass National Approval 🚀

Brief:

• Tesla's xBot customer service, Volvo's "Xiaowo" intelligent assistant, and Mercedes-Benz's virtual assistant have successfully passed national filing or approval. They are among the first batch of approved foreign enterprise large model products to go live nationwide, offering users intelligent services such as product consultation and vehicle usage Q&A. 🤖🚗✨

Related Links:

• First Batch Nationwide: Two Foreign Enterprise AI Large Models in Shanghai, from Tesla and Volvo, Pass National Approval

微软联网多个数据中心，构建虚拟AI超级计算机

简报：

• 微软于当地时间11月12日宣布，已着手构建其首个“AI超级工厂”。
• 该工厂由多个新一代Fairwater数据中心站点组成，并通过新型AI WAN专用网络互联。
• 整个分布式工厂将作为一台虚拟超级计算机运行，整合数十万颗GPU、EB级存储和数百万CPU核心。
• 其目标是支持数千亿参数级别的大型AI模型训练，解决单个数据中心无法处理的前沿问题。
• 目前位于美国威斯康星州和佐治亚州的两座Fairwater数据中心已投入运营。

相关链接：

• 微软打造首个 AI 超级工厂：多座新一代 Fairwater 数据中心构成站点网络

Microsoft Interconnects Data Centers to Build Virtual AI Supercomputer

Brief:

• Microsoft announced on November 12 local time that it has begun building its first "AI superfactory." 🚀
• This factory comprises multiple next-generation Fairwater data center sites, interconnected via a new AI WAN dedicated network.
• The entire distributed factory will operate as a virtual supercomputer, integrating hundreds of thousands of GPUs, exabyte-scale storage, and millions of CPU cores. ✨
• Its goal is to support the training of large-scale AI models with hundreds of billions of parameters, tackling frontier problems that single data centers cannot handle.
• Currently, two Fairwater data centers in Wisconsin and Georgia, USA, are already operational. 🌐

Related Links:

• 微软打造首个 AI 超级工厂：多座新一代 Fairwater 数据中心构成站点网络

Tachyum公布2nm通用处理器Prodigy，宣称AI性能远超英伟达

简报：

• 通用处理器企业Tachyum公布了其2nm Prodigy通用处理器的核心规格，宣称是首款推理性能突破1000 PFLOPs的芯片。
• Prodigy处理器的顶级型号Prodigy Ultimate拥有高达1024个内核，频率可达6.0GHz，支持24通道DDR5内存。
• Tachyum声称，基于其Prodigy Ultimate构建的AI机架系统，在推理性能上可达到英伟达Vera Rubin平台的22.3倍。
• 该公司表示，在2.2亿美元投资的支持下，2nm Prodigy芯片正准备进行流片。

相关链接：

• Tachyum 晒 2nm Prodigy 通用处理器设计：宣称推理性能达英伟达 Rubin 平台 20+ 倍

Tachyum Unveils 2nm General-Purpose Prodigy Processor, Claiming AI Performance Far Exceeding NVIDIA 🚀

Brief:

• Universal processor company Tachyum has announced the core specifications of its 2nm Prodigy general-purpose processor, claiming it's the first chip to break through 1000 PFLOPs in inference performance.
• The top-tier Prodigy Ultimate model of the Prodigy processor features up to 1024 cores, clocked at up to 6.0GHz, and supports 24-channel DDR5 memory.
• Tachyum asserts that an AI rack system built with its Prodigy Ultimate can achieve 22.3 times the inference performance of NVIDIA's Vera Rubin platform. 💡
• The company states that, backed by $220 million in investment, the 2nm Prodigy chip is preparing for tape-out. 🔥

Related Link:

• Tachyum Unveils 2nm Prodigy General-Purpose Processor Design: Claims 20+ Times Inference Performance Over NVIDIA's Rubin Platform

宇树发布轮式人形机器人G1-D，主打数据采集与训练

简报：

• 宇树科技于11月13日推出一款名为G1-D的轮式人形机器人，该机器人是其全新人形机器人数采训练全栈解决方案的核心。
• G1-D分为通用版和旗舰版，身高范围为1260-1680mm，头部配备高清双目相机，手部也装有高清相机。
• 旗舰版可选配移动底盘，实现最高1.5m/s的移动速度；整机自由度最高达19个（不含末端），单臂最大负载约3kg。
• 该机器人采用轮式与升降结合的设计，最高续航约6小时，并可选配二指至五指等多种灵巧手。

相关链接：

• 宇树推出首款轮式人形机器人 G1-D：手上装摄像头，可选配移动底盘

Unitree Unveils G1-D Wheeled Humanoid Robot, Specializing in Data Collection and Training

Brief:

• Unitree Robotics launched its G1-D wheeled humanoid robot on November 13, serving as the core of its new full-stack humanoid robot data collection and training solution. 🤖
• The G1-D comes in General and Flagship editions, boasting a height range of 1260-1680mm. Its head is equipped with high-definition binocular cameras, and its hands also feature HD cameras. 📸
• The Flagship edition offers an optional mobile chassis, enabling a top movement speed of 1.5m/s. The entire machine features up to 19 degrees of freedom (excluding end effectors), with a maximum single-arm load capacity of approximately 3kg.
• The robot features a hybrid design combining wheels with an elevating mechanism, offering a maximum battery life of approximately 6 hours. It can also be optionally equipped with various dexterous hands, ranging from two-finger to five-finger configurations. ✨

Related Links:

• Unitree Launches Its First Wheeled Humanoid Robot G1-D: Cameras on Hands, Optional Mobile Chassis