Nvidia 把 AI 智能体从云端拉到本地 PC，做边缘计算或桌面 AI 应用的开发者可以直接关注 RTX Spark；Vera CPU 专为智能体设计，做数据中心或高频交易的团队值得研究。
Nvidia 在 COMPUTEX 上发布多项重磅产品，覆盖 AI 计算的不同层面。RTX Spark 是与微软合作的新一代 AI 超算芯片，可直接在 Windows 电脑上运行 AI 智能体，无需云端，今年秋季发货。Vera 是 Nvidia 首款专为 AI 智能体设计的 CPU，性能比竞品快 1.8 倍，纽约证券交易所计划用其扩容。Cosmos 3 是开源世界模型，可让机器人和自动驾驶汽车预测场景，现已可用。Nemotron 3 Ultra 是 550B 参数的开源权重模型，本周发布，对标 Kimi K2.6、GLM 5.1 等。
Nvidia just made some major moves at COMPUTEX, hitting several different layers of computing. Some ...
Nvidia just made some major moves at COMPUTEX, hitting several different layers of computing. Some of the biggest annoucements: 1. RTX Spark is a new class of AI superchips NVIDIA built with Microsoft to run AI agents directly on Windows computers instead of in the cloud. Nvidia describes the shift as going "from tool to teammate" and "the first Windows computers purpose-built for personal agents." Built-in support for agents like Hermes and OpenClaw. Shipping this fall. 2. Vera, what Nvidia calls "the first CPU built for AI agents." 1.8x faster than rival CPU options. Jensen Huang: "AI agents will be the largest users of computing. Vera is the first CPU designed for that future — built to run agentic AI at hyperscale with extraordinary performance, efficiency and programmability." The New York Stock Exchange, which moves more than 1.1T messages a day, plans to use it to scale capacity. Also available this fall. 3. Cosmos 3, an open AI model that lets a robot or self-driving car understand a scene and predict what's about to happen before it moves. Can be used for vision tasks, as a world model simulator, or for robotics training. Unlike the other two, Cosmos 3 is already available now. 4. Nemotron 3 Ultra - Nvidia's new 550B parameter open weight model coming this week. Tops U.S. open models and direclty competes with top options lke Kimi K2.6, GLM 5.1, Qwen 3.5 and others. Your browser does not support the video tag. 🔗 View on Twitter 💬 2 🔄 0 ❤️ 10 👀 907 📊 3 ⚡