近期， Vibe Coding 公司「 InferNet 」 宣布完成天使轮融资，由锦秋基金独家投资，澜松资本担任本轮以及后续融资独家财务顾问。 创始团队由 99 年创始人与 97 年 CTO 组成，其前作 NextChat （原 ChatGPT-Next-Web ）在 GitHub 获得 85.5k Stars ，成为最受欢迎的 LLM 开源 UI 之 一，并于 2024 年被市场上最炙手可热的 AI 公司「 Manus 」成功收购。 NextChat （ ChatGPT-Next-Web ）是一款轻量级、跨平台的开源聊天界面应用，它支持通过一键部署方式在 Vercel 等平台上快速搭建私人 AI 聊天工具。 其客户端非常小（约 5 MB ），支持 Web 、 Windows 、 macOS 、 Linux 等多个平台。项目在 GitHub 获得 85.5k Star / 61.1k Fork / 260+ 贡献者，是 LLM 开源 UI 的首选。 NextChat （ ChatGPT-Next-Web ） 从 " 聊天界面 " 到 " 全栈引擎 "： Claude 3 Sonnet 的发布让 AI 生成的 ...

来源：china.com.cn 演讲指出，当前AI行业呈现"智算哑铃态"发展趋势，应用场景与算力需求双向爆发，而基础模型架构逐 步收敛。在这一背景下，浙江移动依托华为昇腾AI基础软硬件平台，构建起覆盖算法开发、训练加 速、推理部署的全栈国产化智算能力。 5月17日，2025世界电信和社会信息纪念日活动于江西南昌举办，浙移数科副总经理汤人杰出席AI与行 业大模型研讨会，发表题为《浙江移动大模型全栈工程交付能力最佳实践》的主题演讲，深入剖析AI 技术发展趋势与产业实践路径，并重点分享了浙江移动在智算国产化与AI工程化领域的创新成果。 1.智算产业升级加速，国产化布局抢占先机 在实践层面，成功落地xx海关、xx人民医院、xx电力等覆盖政务、医疗、能源多个领域的标杆案例，充 分彰显技术赋能的价值。 ①xx海关项目基于昇腾AI算力底座原生开发AI政务助手应用，提供智能问答、公文助手、智能检索等 能力，项目实践过程中针对RAG Rerank多阶段协同困难、多阶段协同困难、效果评估复杂度高三大技 术挑战，通过入库文档精细化知识处理、问题段落重排算法(QPR)和Group-Weighted Rerank算法进行优 化实践； ...

Core Viewpoint - Lepton AI, founded by former Alibaba executives, is reportedly in talks to be acquired by NVIDIA for an estimated amount in the hundreds of millions of dollars, focusing on AI cloud services and GPU rental business [1][7]. Company Overview - Lepton AI was established in 2023 and raised $11 million in seed funding, aiming to provide an efficient and scalable AI application platform that allows developers with minimal AI knowledge to deploy AI models with just 2-3 lines of code [2]. - The founding team includes notable figures from Alibaba and Meta, such as former Alibaba VP Jia Yangqing and AI researchers Bai Junjie and Li Xiang [3][4]. Strategic Importance - NVIDIA is interested in Lepton AI primarily for its cloud service capabilities, which could help diversify its business and reduce reliance on major clients like Amazon and Google, who are developing their own chips [6][7]. - Lepton AI is the first company NVIDIA is directly acquiring, indicating a strategic move towards enhancing its cloud computing offerings [8]. Technological Contributions - Lepton AI has gained recognition for its open-source contributions, including a demonstration of a conversational search engine implemented with only 500 lines of code [11]. - The company has invested approximately $100 million in Coreweave for high-end GPU supply and participated in significant funding rounds for AI cloud providers, including $480 million in Lambda's Series D and $305 million in TogetherAI's Series B [12]. Business Model - Lepton AI does not own any GPUs but provides a machine learning platform for managing and monitoring GPU health, allowing users to rent GPUs from other suppliers or purchase support and platform access [18][19]. - By simplifying AI training and deployment, Lepton AI enables users to focus on product development rather than technical complexities [21].