目录
单词学习：![[00638a80-da4d-488e-81ee-f9669b9d7a4d.mp3]]
新闻播报：![[3a70fc58-e0fc-4164-9922-b8b5dfd50fc0.mp3]]
蚂蚁灵光AI助手下载量火爆，四天破百万
简报：
- 蚂蚁集团于11月18日推出全模态通用AI助手“灵光”，上线4天下载量即突破100万。
- 其百万下载速度超过了ChatGPT、Sora2等全球主流AI产品，并冲上App Store中国区免费榜第六位。
- “灵光”是业内首个支持全代码生成多模态内容的AI助手，核心功能包括“自然语言30秒生成小应用”。
- 由于用户需求旺盛，上线4天内官方已紧急进行8轮服务器扩容。
相关链接：
Ant Lingguang AI Assistant Downloads Go Viral, Surpassing One Million in Four Days 🚀
Brief:
- Ant Group launched its full-modal general AI assistant, "Lingguang," on November 18th, with downloads exceeding 1 million within just four days of its release.
- This rapid pace of 1 million downloads outstripped global mainstream AI products like ChatGPT and Sora2, and propelled "Lingguang" to the sixth spot on the App Store China's free app chart. ⭐
- "Lingguang" is the industry's first AI assistant to support full-code generation of multimodal content, with a core feature including "generating mini-apps from natural language in 30 seconds." 🧠
- Due to robust user demand, the official team has urgently undertaken 8 rounds of server capacity expansion within four days of its launch.
Related Links:
腾讯发布HunyuanVideo 1.5，大幅降低AI视频生成门槛
简报：
- 腾讯混元大模型团队正式发布全新视频生成模型HunyuanVideo 1.5，该模型基于Diffusion Transformer（DiT）架构，参数量为8.3B。
- 模型支持通过文字或图文结合的方式，生成5至10秒的高清视频，并兼容中英文输入。
- HunyuanVideo 1.5显著降低了技术门槛，仅需14G显存的消费级显卡即可运行，远低于此前开源模型的要求。
- 目前，该模型已在腾讯“元宝”平台上线供公众体验，并已在Hugging Face和GitHub上开源。
相关链接：
Tencent Releases HunyuanVideo 1.5, Significantly Lowering AI Video Generation Barrier
Briefing:
- Tencent's Hunyuan LLM team officially released the new video generation model, HunyuanVideo 1.5. This model is based on the Diffusion Transformer (DiT) architecture and boasts 8.3 billion parameters.
- The model supports generating 5-10 second high-definition videos via text or a combination of text and images, and is compatible with both Chinese and English input. 🚀
- HunyuanVideo 1.5 significantly lowers the technical barrier, requiring only a consumer-grade graphics card with 14GB of VRAM to run, far less than what previous open-source models demanded. ✨
- Currently, the model is available for public experience on Tencent's 'Yuanbao' platform and has been open-sourced on Hugging Face and GitHub. 💡
Related Links:
传美国考虑允许英伟达对华出售H200芯片
简报：
- 据路透社21日援引消息人士透露，特朗普政府正考虑批准向中国出口芯片制造商英伟达的H200人工智能芯片。
- 报道称，负责监管出口管制的美国商务部正在审查对华出口限制的调整，但相关计划可能会发生变动。
- 英伟达尚未就此消息直接置评。
相关链接：
US Reportedly Considers Allowing Nvidia to Sell H200 Chips to China 🤔
Brief:
- According to Reuters on the 21st, citing sources, the Trump administration is reportedly considering approving the export of chipmaker Nvidia's H200 artificial intelligence chips to China.
- The report states that the U.S. Commerce Department, which oversees export controls, is reviewing adjustments to export restrictions on China, but the relevant plans could change.
- Nvidia has not yet directly commented on this news. 💾
Related Link:
OpenAI发布GPT-5，聚焦加速数学与科学领域研究
简报：
- OpenAI近日宣布推出其最新的GPT-5模型，该模型具备更强的计算能力和语言理解能力，能够更准确地理解和生成数学公式及科学概念。
- 在药物发现领域，GPT-5可通过分析大量生物医学文献和数据，快速识别潜在的药物分子和疗法，从而加速临床试验进程。
- 在新材料研究方面，该模型能帮助科研人员借助其数据处理能力，模拟和预测材料特性，以加速新材料的研发过程。
相关链接：
OpenAI Launches GPT-5, Focusing on Accelerating Research in Mathematics and Science
Brief:
- OpenAI recently announced the launch of its latest GPT-5 model, which boasts enhanced computational power and language understanding capabilities, enabling it to more accurately comprehend and generate mathematical formulas and scientific concepts. 🚀
- In the field of drug discovery, GPT-5 can rapidly identify potential drug molecules and therapies by analyzing vast amounts of biomedical literature and data, thereby accelerating the clinical trial process. 💊
- For new material research, the model's data processing capabilities can assist researchers in simulating and predicting material properties, thus speeding up the development of novel materials. 🔬
Related Links:
- OpenAI Releases New GPT-5 Model to Accelerate Math and Science Research
- OpenAI Launches "ChatGPT for Teachers" to Empower K-12 Education
小米开源具身智能大模型，融合自动驾驶与机器人技术
简报：
- 小米于今日正式发布并宣布全面开源其具身智能大模型MiMo-Embodied，旨在打通自动驾驶与具身智能两大领域。
- 该模型实现了任务统一建模，具备跨域能力覆盖，同步支持具身智能（可供性推理、任务规划、空间理解）与自动驾驶（环境感知、状态预测、驾驶规划）的核心任务。
- MiMo-Embodied采用“具身/自驾能力学习、CoT推理增强、RL精细强化”的多阶段训练策略，提升了真实环境部署的可靠性。
- 在29项核心基准测试中，该模型性能全面优于现有开源、闭源及专用模型，在具身智能和自动驾驶领域的多项测试上取得SOTA（State-of-the-art）成绩。
相关链接：
Xiaomi Open-Sources Embodied AI Large Model, Integrating Autonomous Driving and Robotics Technology 🤖
Briefing:
- Xiaomi today officially launched and announced the full open-sourcing of its embodied intelligent large model, MiMo-Embodied, aiming to bridge the two major fields of autonomous driving and embodied intelligence. 💡
- The model achieves unified task modeling and possesses cross-domain capability coverage, simultaneously supporting core tasks in embodied AI (affordance reasoning, task planning, spatial understanding) and autonomous driving (environmental perception, state prediction, driving planning).
- MiMo-Embodied adopts a multi-stage training strategy comprising "embodiment/self-driving capability learning, CoT (Chain of Thought) reasoning enhancement, and RL (Reinforcement Learning) refined reinforcement," enhancing its reliability for real-world deployment.
- In 29 core benchmark tests, the model's performance comprehensively surpasses existing open-source, closed-source, and specialized models, achieving SOTA (State-of-the-art) results in multiple tests across both embodied AI and autonomous driving domains. 🚀
Related Links:
- Xiaomi Launches and Open-Sources Embodied Large Model MiMo-Embodied
- Xiaomi Auto Responds to 'Late Start' Criticism in Intelligent Driving: AI Investment to Exceed 7 Billion by 2025
- Xiaomi Smart Band 10 / AI Glasses New Versions Begin Gray Release, Full Rollout Expected Next Week
小米回应智驾起步晚：2025年将投超70亿加码AI
简报：
- 在2025广州车展上，小米汽车副总裁李肖爽针对外界关于“小米辅助驾驶起步晚”的质疑进行了回应。
- 李肖爽表示辅助驾驶的核心本质是AI技术，并透露为支撑该战略，小米在2025年在AI领域的研发投入将超过70亿元。
- 目前，小米智能驾驶团队规模已突破1800人，拥有108名博士，并已设立北京、上海、武汉三大研发中心。
相关链接：
Xiaomi Addresses 'Late Start' in Intelligent Driving, Pledges Over ¥7 Billion for AI by 2025 🚗
Briefing:
- At the 2025 Guangzhou Auto Show, Li Xiaoshuang, Vice President of Xiaomi Auto, responded to external skepticism regarding Xiaomi's 'late start' in assisted driving.
- Li Xiaoshuang stated that the core essence of assisted driving is AI technology 🧠, revealing that to support this strategy, Xiaomi's R&D investment in AI will exceed ¥7 billion in 2025.
- Currently, Xiaomi's intelligent driving team has grown to over 1,800 people, including 108 PhD holders, and has established three major R&D centers in Beijing, Shanghai, and Wuhan. 🚀
Related Links:
- Xiaomi Auto Responds to 'Late Start' in Smart Driving Query: Over ¥7 Billion in AI Investment by 2025
- Xiaomi Releases and Open-Sources its Embodied Large Model MiMo-Embodied
- New Versions of Xiaomi Band 10 / AI Glasses Have Started Grey-Scale Testing, Full Rollout Expected Next Week
谷歌发布Nano Banana Pro图像模型，支持4K与联网生成
简报：谷歌发布了基于Gemini 3的新一代图像生成模型Nano Banana Pro，该模型支持高达4K分辨率、联网搜索和SynthID/C2PA双重水印验证，并已集成至NotebookLM等产品中，其定价也较前代上涨3至6倍。
相关链接：
Google Launches Nano Banana Pro Image Model, Supporting 4K and Online Generation 🚀
Briefing: Google has launched Nano Banana Pro, a new generation image generation model based on Gemini 3. This model supports up to 4K resolution, online search, and dual watermarking verification (SynthID/C2PA). It has been integrated into products like NotebookLM, with its pricing increasing 3 to 6 times compared to its predecessor. ✨
Related Links: 🔗
- Google Launches Nano Banana Pro: Based on Gemini 3, Supporting 4K, Online Search, and C2PA Verification
- Google NotebookLM Major Update: AI Generates Professional PPTs with One Click, Stunning Images Surpass Gemini App!
如果对你有用的话，可以打赏哦
打赏
本文作者:topwind
本文链接:
版权声明:本博客所有文章除特别声明外，均采用 BY-NC-SA 许可协议。转载请注明出处！
目录