{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-23T07:21:26.498Z","headline":"NVIDIA 发布 Cosmos 3 Edge：40 亿参数开放世界模型，可在设备端推理并生成机器人动作","description":"NVIDIA 发布 Cosmos 3 Edge，一款 40 亿参数的开放世界模型，专为设备端运行设计。该模型可帮助机器人和视觉 AI 智能体理解环境、实时推理并在本地生成机器人动作。Cosmos 3 系列此前已于 2026 年 5 月 31 日在 GTC Taipei 推出 Cosmos 3 Nano（160 亿参数）和 Cosmos 3 Super（640 亿参数）。","url":"https://www.aioga.com/news/cmrucw21u0xgubi7fmmu88bby/","mainEntityOfPage":"https://www.aioga.com/news/cmrucw21u0xgubi7fmmu88bby/","datePublished":"2026-07-21T07:48:28.000Z","dateModified":"2026-07-21T07:48:28.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://www.marktechpost.com/2026/07/21/nvidia-releases-cosmos-3-edge-a-4b-parameter-open-world-model-that-reasons-and-generates-robot-actions-on-device","https://aihot.virxact.com/items/cmrucw21u0xgubi7fmmu88bby"],"canonicalUrl":"https://www.aioga.com/news/cmrucw21u0xgubi7fmmu88bby/","directAnswer":{"@type":"Answer","text":"Aioga 编辑摘要：NVIDIA 发布 Cosmos 3 Edge，一款 40 亿参数的开放世界模型，专为设备端运行设计。 Aioga 将其归入「模型更新」方向，重点关注它对真实使用和行业竞争的影响。","url":"https://www.aioga.com/news/cmrucw21u0xgubi7fmmu88bby/","dateCreated":"2026-07-21T07:48:28.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"marktechpost.com source article","url":"https://www.marktechpost.com/2026/07/21/nvidia-releases-cosmos-3-edge-a-4b-parameter-open-world-model-that-reasons-and-generates-robot-actions-on-device","datePublished":"2026-07-21T07:48:28.000Z","provider":{"@type":"Organization","name":"marktechpost.com","url":"https://www.marktechpost.com/2026/07/21/nvidia-releases-cosmos-3-edge-a-4b-parameter-open-world-model-that-reasons-and-generates-robot-actions-on-device"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmrucw21u0xgubi7fmmu88bby","datePublished":"2026-07-21T07:48:28.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmrucw21u0xgubi7fmmu88bby"}}],"aggregationSource":"MarkTechPost（RSS）","originalPublisher":{"name":"marktechpost.com","url":"https://www.marktechpost.com/2026/07/21/nvidia-releases-cosmos-3-edge-a-4b-parameter-open-world-model-that-reasons-and-generates-robot-actions-on-device"},"article":{"id":"cmrucw21u0xgubi7fmmu88bby","slug":"cmrucw21u0xgubi7fmmu88bby","url":"https://www.aioga.com/news/cmrucw21u0xgubi7fmmu88bby/","title":"NVIDIA 发布 Cosmos 3 Edge：40 亿参数开放世界模型，可在设备端推理并生成机器人动作","title_en":"NVIDIA Releases Cosmos 3 Edge： A 4B-Parameter Open World Model That Reasons and Generates Robot Actions On-Device","summary":"NVIDIA 发布 Cosmos 3 Edge，一款 40 亿参数的开放世界模型，专为设备端运行设计。该模型可帮助机器人和视觉 AI 智能体理解环境、实时推理并在本地生成机器人动作。Cosmos 3 系列此前已于 2026 年 5 月 31 日在 GTC Taipei 推出 Cosmos 3 Nano（160 亿参数）和 Cosmos 3 Super（640 亿参数）。","source":"MarkTechPost（RSS）","sourceUrl":"https://www.marktechpost.com/2026/07/21/nvidia-releases-cosmos-3-edge-a-4b-parameter-open-world-model-that-reasons-and-generates-robot-actions-on-device","aiHotUrl":"https://aihot.virxact.com/items/cmrucw21u0xgubi7fmmu88bby","publishedAt":"2026-07-21T07:48:28.000Z","category":"模型更新","score":56,"selected":false,"articleBody":["NVIDIA has released Cosmos 3 Edge ：https://huggingface.co/blog/nvidia/cosmos3edge, a 4-billion-parameter open world model built to run on-device. It helps robots and vision AI agents understand surroundings, reason in real time, and generate robot actions locally.","The Cosmos 3 family：https://huggingface.co/collections/nvidia/cosmos3 included Cosmos 3 Nano (16B) and Cosmos 3 Super (64B) shipped on May 31, 2026 at GTC Taipei：https://www.marktechpost.com/2026/06/03/nvidia-releases-cosmos-3-a-two-tower-mixture-of-transformers-foundation-model-unifying-physical-reasoning-world-generation-and-action-generation/. Edge is the third and smallest tier, at roughly one-sixteenth the size of Super.","The problem is specific. Machines operate at the edge in factories, warehouses, and hospitals. They need data center–level performance on memory-constrained systems. Cosmos 3 Edge targets that gap.","A world model learns how an environment changes over time. It represents objects, motion, spatial relationships, and the effects of actions.","Consider a robot reaching for an object. Recognizing the object is only the first step. The robot must also track where the object is, how its gripper moves, and what happens on contact. A world model reasons about these relationships. It can predict the visual result of an action, infer the action that caused a change, or generate an action to reach a goal.","Cosmos 3 Edge brings these capabilities into one on-device model. Its shared representation lets a system understand the current world state, simulate possible futures, and connect those futures to actions.","Cosmos 3 uses a Mixture-of-Transformers architecture with two towers, described in the NVIDIA’s technical report：https://research.nvidia.com/labs/cosmos-lab/cosmos3/technical-report.pdf.","The autoregressive tower processes vision and text tokens for understanding and reasoning. The diffusion tower processes vision, audio, and action tokens for prediction, generation, and neural simulation.","The two towers keep separate normalization layers and multilayer perceptrons. They share multimodal attention layers, which align information across language, video, audio, and action. This lets the model reason about a scene before it generates an output.","The attention pattern adapts to each modality. Language uses causal attention, where each token attends to earlier tokens. Diffusion tokens attend more broadly to the available context, supporting coherent prediction and generation. Depending on the task, the model emits reasoning tokens from the autoregressive tower, or denoised video and action tokens from the diffusion tower.","Cosmos 3 Edge uses a 2B dense transformer for its reasoner, and follows Qwen3-VL-compatible message conventions for image and video inputs, per the Cosmos GitHub repository：https://github.com/nvidia/cosmos.","Physical systems describe actions differently. A vehicle uses ego pose and movement. A camera uses camera motion. A robot arm uses the pose of its end effector, and a gripper adds grasp state.","Cosmos 3 maps these embodiments into a common action representation. Actions are encoded as compact geometric vectors that capture translation, rotation, and manipulation state.","This connects control to the visual structure of the world. The model associates pixel changes with physical motion and control inputs. Generated video then becomes more than a prediction. It represents how the world should change in response to an action.","Supported action dimensions depend on the embodiment. The Cosmos GitHub repository：https://github.com/nvidia/cosmos lists camera motion (9D), autonomous vehicle (9D), egocentric motion (57D), single-arm robot (10D), dual-arm robot (20D), and humanoid robot (29D).","As a policy, Cosmos 3 Edge predicts an action together with its expected visual consequence. Current state goes in; an action and its likely visual outcome come out.","Action flows in both directions. The model can predict the effect of an action, or infer the action from its effect. This connects world modeling directly to robot policy training and evaluation.","NVIDIA also released Cosmos 3 Edge Policy (DROID) . It is a robot manipulation policy post-trained on the DROID dataset for pick-and-place tasks, with post-training scripts included. Developers can fine-tune on a small H100 cluster or an NVIDIA DGX Station before deployment.","Cosmos 3 Edge delivers memory-efficient inference across NVIDIA edge computers. Targets include NVIDIA RTX PRO GPUs, NVIDIA DGX, GeForce RTX GPUs, and NVIDIA Jetson：https://blogs.nvidia.com/blog/jetson-thor-robotics-edge-ai-agent/, including the newly announced Jetson T2000 and T3000 modules.","As a post-trained world action model (WAM), the model operates at robot-control resolution of 640×360 observations. On NVIDIA Jetson Thor it generates 32 actions per inference, while achieving real-time control at 15 Hz. For generation, the Edge tier supports 256p and 480p resolutions, 12–30 fps, and 50–150 frames.","Using the open Cosmos framework, developers can post-train Cosmos 3 Edge for a specific embodiment and sensor set in about a day. NVIDIA positions a GeForce RTX 3070 or better as a local on-ramp for prototyping.","Sources: Hugging Face launch post：https://huggingface.co/blog/nvidia/cosmos3edge, Cosmos3-Edge model card：https://huggingface.co/nvidia/Cosmos3-Edge, Cosmos 3 collection：https://huggingface.co/collections/nvidia/cosmos3, NVIDIA technical report (PDF)：https://research.nvidia.com/labs/cosmos-lab/cosmos3/technical-report.pdf, Cosmos GitHub：https://github.com/nvidia/cosmos, NVIDIA developer blog：https://developer.nvidia.com/blog/develop-physical-ai-reasoning-world-and-action-models-with-nvidia-cosmos-3/, Jetson Thor blog：https://blogs.nvidia.com/blog/jetson-thor-robotics-edge-ai-agent/ and NVIDIA Newsroom: Japan coalition：https://nvidianews.nvidia.com/news/japans-robotics-and-manufacturing-leaders-build-on-nvidia-cosmos-to-advance-physical-ai-frontier","Asif Razzaq is the CEO of Marktechpost Media Inc.. As a visionary entrepreneur and engineer, Asif is committed to harnessing the potential of Artificial Intelligence for social good. His most recent endeavor is the launch of an Artificial Intelligence Media Platform, Marktechpost, which stands out for its in-depth coverage of machine learning and deep learning news that is both technically sound and easily understandable by a wide audience. The platform boasts of over 2 million monthly views, illustrating its popularity among audiences.","Build an Agentic Event Venue Operator [Full Codes]：https://pxllnk.co/twdn5","Thanks! Our team will contact you soon 🙌"],"articleImages":[{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2019/06/Screen-Shot-2021-09-14-at-9.02.24-AM-300x300.png","alt":"","afterParagraph":21,"url":"/media/articles/cmrucw21u0xgubi7fmmu88bby/787a6d54564e8e19.webp"},{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2026/07/blog19132-1-8-100x70.png","alt":"Google Releases Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber","afterParagraph":22,"url":"/media/articles/cmrucw21u0xgubi7fmmu88bby/39896896ab8c8014.webp"}],"mediaStatus":"ok","articleBodyZh":["英伟达发布了 Cosmos 3 Edge：https://huggingface.co/blog/nvidia/cosmos3edge，这是一款拥有 40 亿参数的开放世界模型，旨在在设备端运行。它帮助机器人和视觉 AI 代理理解周围环境，实时推理，并在本地生成机器人动作。","Cosmos 3 系列：https://huggingface.co/collections/nvidia/cosmos3 包括 Cosmos 3 Nano（16B）和 Cosmos 3 Super（64B），于 2026 年 5 月 31 日在 GTC 台北发布：https://www.marktechpost.com/2026/06/03/nvidia-releases-cosmos-3-a-two-tower-mixture-of-transformers-foundation-model-unifying-physical-reasoning-world-generation-and-action-generation/。Edge 是第三个也是最小的等级，大约是 Super 的六分之一大小。","问题是具体的。机器在工厂、仓库和医院的边缘运行。它们需要在内存受限的系统上达到数据中心级别的性能。Cosmos 3 Edge 针对这一空白。","世界模型学习环境随时间变化的方式。它表示物体、运动、空间关系以及动作的影响。","考虑一个机器人去抓取一个物体。识别物体只是第一步。机器人还必须跟踪物体的位置、夹爪的运动以及接触时发生的情况。世界模型会对这些关系进行推理。它可以预测动作的视觉结果，推断导致变化的动作，或生成达成目标的动作。","Cosmos 3 Edge 将这些能力整合到一个设备端模型中。它的共享表示使系统能够理解当前的世界状态、模拟可能的未来，并将这些未来与动作连接起来。","Cosmos 3 使用具有两座塔的 Mixture-of-Transformers 架构，详见英伟达的技术报告：https://research.nvidia.com/labs/cosmos-lab/cosmos3/technical-report.pdf。","自回归塔处理视觉和文本标记，以实现理解和推理。扩散塔处理视觉、音频和动作标记，用于预测、生成和神经模拟。","这两座塔保持独立的归一化层和多层感知器。它们共享多模态注意力层，用于跨语言、视频、音频和动作对齐信息。这使模型能够在生成输出之前对场景进行推理。","注意力模式会根据不同模态进行调整。语言使用因果注意力，每个 token 关注之前的 token。扩散 token 则会更广泛地关注可用上下文，以支持连贯的预测和生成。根据任务的不同，模型会从自回归塔发出推理 token，或者从扩散塔发出去噪视频和动作 token。","Cosmos 3 Edge 的推理器使用一个 2B 的密集变换器，并按照 Cosmos GitHub 仓库中的说明，遵循 Qwen3-VL 兼容的图像和视频输入消息约定：https://github.com/nvidia/cosmos。","物理系统对动作的描述各不相同。车辆使用自身姿态和运动。相机使用相机运动。机械臂使用末端执行器的姿态，而夹爪则添加抓取状态。","Cosmos 3 将这些体现形式映射到一个通用的动作表示中。动作被编码为紧凑的几何向量，捕捉平移、旋转和操作状态。","这将控制与世界的视觉结构连接起来。模型将像素变化与物理运动和控制输入相关联。生成的视频因此不仅仅是预测，它表示世界应如何响应动作而发生变化。","支持的动作维度取决于体现形式。Cosmos GitHub 仓库：https://github.com/nvidia/cosmos 列出了相机运动（9D）、自动驾驶车辆（9D）、自我中心运动（57D）、单臂机器人（10D）、双臂机器人（20D）和类人机器人（29D）。","作为一种策略，Cosmos 3 Edge 会同时预测动作及其预期的视觉结果。当前状态输入；输出为动作及其可能的视觉结果。","动作的流动是双向的。模型可以预测某个动作的效果，或者从效果中推断动作。这将世界建模直接与机器人策略训练和评估连接起来。","NVIDIA 还发布了 Cosmos 3 Edge Policy（DROID）。它是一种机器人操作策略，在 DROID 数据集上进行了取放任务的后训练，并包含后训练脚本。开发者可以在小型 H100 集群或 NVIDIA DGX Station 上进行微调，然后再部署。","Cosmos 3 Edge 在 NVIDIA 边缘计算机上提供高效内存推理。目标设备包括 NVIDIA RTX PRO GPU、NVIDIA DGX、GeForce RTX GPU 以及 NVIDIA Jetson：https://blogs.nvidia.com/blog/jetson-thor-robotics-edge-ai-agent/，包括新发布的 Jetson T2000 和 T3000 模块。","作为一个后训练世界动作模型（WAM），该模型以 640×360 分辨率的机器人控制观测进行运行。在 NVIDIA Jetson Thor 上，它每次推理生成 32 个动作，同时在 15 Hz 实现实时控制。对于生成任务，Edge 级别支持 256p 和 480p 分辨率，帧率为 12–30 fps，以及 50–150 帧。","使用开放的 Cosmos 框架，开发者可以在大约一天内针对特定的实体和传感器集完成 Cosmos 3 Edge 的后训练。NVIDIA 建议 GeForce RTX 3070 或更高版本作为原型设计的本地入门设备。","来源：Hugging Face 发布文章：https://huggingface.co/blog/nvidia/cosmos3edge, Cosmos3-Edge 模型卡：https://huggingface.co/nvidia/Cosmos3-Edge, Cosmos 3 系列：https://huggingface.co/collections/nvidia/cosmos3, NVIDIA 技术报告 (PDF)：https://research.nvidia.com/labs/cosmos-lab/cosmos3/technical-report.pdf, Cosmos GitHub：https://github.com/nvidia/cosmos, NVIDIA 开发者博客：https://developer.nvidia.com/blog/develop-physical-ai-reasoning-world-and-action-models-with-nvidia-cosmos-3/, Jetson Thor 博客：https://blogs.nvidia.com/blog/jetson-thor-robotics-edge-ai-agent/ 以及 NVIDIA 新闻中心：日本联盟：https://nvidianews.nvidia.com/news/japans-robotics-and-manufacturing-leaders-build-on-nvidia-cosmos-to-advance-physical-ai-frontier","Asif Razzaq 是 Marktechpost Media Inc. 的首席执行官。作为一位有远见的企业家和工程师，Asif 致力于利用人工智能的潜力造福社会。他最近的项目是推出人工智能媒体平台 Marktechpost，该平台以对机器学习和深度学习新闻的深入报道而著称，内容既技术可靠又易于广大受众理解。该平台每月浏览量超过 200 万次，显示了其受欢迎程度。","构建一个能动事件场馆运营商 [完整代码]：https://pxllnk.co/twdn5","谢谢！我们的团队将很快与您联系 🙌"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Aioga 编辑摘要：NVIDIA 发布 Cosmos 3 Edge，一款 40 亿参数的开放世界模型，专为设备端运行设计。 Aioga 将其归入「模型更新」方向，重点关注它对真实使用和行业竞争的影响。","background":"背景分析：模型与研究类动态需要结合能力边界、开放方式、成本、可用性和真实任务表现判断，单项指标领先不等于已经形成稳定采用。","viewpoint":"Aioga 判断：这条动态更适合作为行业观察信号，当前信息足以建立线索，但不足以推导长期结论。","implications":"影响分析：对相关团队而言，短期应先核对来源、可用范围和实际成本，再判断是否值得接入或跟进。","nextStep":"后续观察：继续观察官方文档、实际可用性、价格变化、开发者反馈和竞品回应。","evidenceRefs":["title","summary","articleBody"],"confidence":"medium","status":"published","aiGenerated":false,"autoApproved":true,"generatedBy":"rule-safe-fallback","generatedAt":"2026-07-23T07:30:40.500Z","sourceHash":"0140e5c03301ae16","validation":{"passed":true,"mode":"rule-safe-fallback","checks":["schema","length","source-attribution","no-html"]}},"tags":["模型更新","MarkTechPost（RSS）"],"translations":{"zh-CN":{"title":"NVIDIA 发布 Cosmos 3 Edge：40 亿参数开放世界模型，可在设备端推理并生成机器人动作","summary":"NVIDIA 发布 Cosmos 3 Edge，一款 40 亿参数的开放世界模型，专为设备端运行设计。该模型可帮助机器人和视觉 AI 智能体理解环境、实时推理并在本地生成机器人动作。Cosmos 3 系列此前已于 2026 年 5 月 31 日在 GTC Taipei 推出 Cosmos 3 Nano（160 亿参数）和 Cosmos 3 Super（640 亿参数）。","category":"模型更新","source":"marktechpost.com","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA 发布 Cosmos 3 Edge：40 亿参数开放世界模型，可在设备端推理并生成机器人动作 - Aioga AI资讯","description":"NVIDIA 发布 Cosmos 3 Edge，一款 40 亿参数的开放世界模型，专为设备端运行设计。该模型可帮助机器人和视觉 AI 智能体理解环境、实时推理并在本地生成机器人动作。Cosmos 3 系列此前已于 2026 年 5 月 31 日在 GTC Taipei 推出 Cosmos 3 Nano（160 亿参数）和 Cosmos 3 Super（640...","url":"https://www.aioga.com/news/cmrucw21u0xgubi7fmmu88bby/"},"en":{"title":"NVIDIA Releases Cosmos 3 Edge: 4 Billion Parameter Open-World Model, Capable of On-Device Inference and Generating Robot Actions","summary":"NVIDIA has released Cosmos 3 Edge, a 4-billion-parameter open-world model designed for on-device operation. This model can help robots and visual AI agents understand their environment, perform real-time reasoning, and generate robotic actions locally. The Cosmos 3 series previously launched Cosmos 3 Nano (16-billion parameters) and Cosmos 3 Super (64-billion parameters) on May 31, 2026, at GTC Taipei.","category":"Models","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA Releases Cosmos 3 Edge: 4 Billion Parameter Open-World Model, Capable of On-Device Inference and Generating Robot Actions - Aioga AI News","description":"NVIDIA has released Cosmos 3 Edge, a 4-billion-parameter open-world model designed for on-device operation. This model can help robots and visual AI agents understand their environ...","url":"https://www.aioga.com/en/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:21:36.622Z"},"ja":{"title":"NVIDIA、Cosmos 3 Edge を発表：40億パラメータのオープンワールドモデルで、デバイス上で推論しロボットの動作を生成可能","summary":"NVIDIA は Cosmos 3 Edge を発表しました。これは 40 億パラメータを持つオープンワールドモデルで、デバイス上での動作を目的として設計されています。このモデルは、ロボットやビジュアル AI エージェントが環境を理解し、リアルタイムで推論を行い、ローカルでロボットの動作を生成するのに役立ちます。Cosmos 3 シリーズはこれ以前に、2026 年 5 月 31 日に GTC Taipei で Cosmos 3 Nano（160 億パラメータ）と Cosmos 3 Super（640 億パラメータ）を発表していました。","category":"モデル更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA、Cosmos 3 Edge を発表：40億パラメータのオープンワールドモデルで、デバイス上で推論しロボットの動作を生成可能 - Aioga AIニュース","description":"NVIDIA は Cosmos 3 Edge を発表しました。これは 40 億パラメータを持つオープンワールドモデルで、デバイス上での動作を目的として設計されています。このモデルは、ロボットやビジュアル AI エージェントが環境を理解し、リアルタイムで推論を行い、ローカルでロボットの動作を生成するのに役立ちます。Cosmos 3 シリーズはこれ以前に、202...","url":"https://www.aioga.com/ja/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:21:48.796Z"},"ko":{"title":"NVIDIA가 Cosmos 3 Edge를 발표했습니다: 40억 매개변수의 오픈 월드 모델로, 장치에서 추론하고 로봇 동작을 생성할 수 있습니다","summary":"NVIDIA는 장치 단에서 실행되도록 설계된 40억 파라미터의 오픈 월드 모델인 Cosmos 3 Edge를 발표했습니다. 이 모델은 로봇과 비전 AI 에이전트가 환경을 이해하고 실시간 추론을 하며 로컬에서 로봇 동작을 생성하는 데 도움을 줄 수 있습니다. Cosmos 3 시리즈는 이전에 2026년 5월 31일 GTC Taipei에서 Cosmos 3 Nano(160억 파라미터)와 Cosmos 3 Super(640억 파라미터)를 출시한 바 있습니다.","category":"모델 업데이트","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA가 Cosmos 3 Edge를 발표했습니다: 40억 매개변수의 오픈 월드 모델로, 장치에서 추론하고 로봇 동작을 생성할 수 있습니다 - Aioga AI 뉴스","description":"NVIDIA는 장치 단에서 실행되도록 설계된 40억 파라미터의 오픈 월드 모델인 Cosmos 3 Edge를 발표했습니다. 이 모델은 로봇과 비전 AI 에이전트가 환경을 이해하고 실시간 추론을 하며 로컬에서 로봇 동작을 생성하는 데 도움을 줄 수 있습니다. Cosmos 3 시리즈는 이전에 2026년 5월 31일 GTC Ta...","url":"https://www.aioga.com/ko/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:22:31.404Z"},"es":{"title":"NVIDIA lanza Cosmos 3 Edge: modelo de mundo abierto de 4.000 millones de parámetros, que puede realizar inferencia en el dispositivo y generar acciones de robots","summary":"NVIDIA lanzó Cosmos 3 Edge, un modelo de mundo abierto con 4 mil millones de parámetros, diseñado para ejecutarse en dispositivos. Este modelo puede ayudar a robots y agentes de IA visual a comprender el entorno, realizar inferencias en tiempo real y generar acciones de robots localmente. La serie Cosmos 3 ya había lanzado el 31 de mayo de 2026 en GTC Taipei el Cosmos 3 Nano (16 mil millones de parámetros) y el Cosmos 3 Super (64 mil millones de parámetros).","category":"Modelos","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA lanza Cosmos 3 Edge: modelo de mundo abierto de 4.000 millones de parámetros, que puede realizar inferencia en el dispositivo y generar acciones de robots - Aioga Noticias de IA","description":"NVIDIA lanzó Cosmos 3 Edge, un modelo de mundo abierto con 4 mil millones de parámetros, diseñado para ejecutarse en dispositivos. Este modelo puede ayudar a robots y agentes de IA...","url":"https://www.aioga.com/es/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:22:29.448Z"},"fr":{"title":"NVIDIA a publié Cosmos 3 Edge : un modèle de monde ouvert à 4 milliards de paramètres, capable d'inférer sur l'appareil et de générer des actions pour les robots","summary":"NVIDIA a lancé Cosmos 3 Edge, un modèle de monde ouvert de 4 milliards de paramètres, conçu pour fonctionner sur les appareils. Ce modèle peut aider les robots et les agents d'IA visuelle à comprendre l'environnement, raisonner en temps réel et générer localement des actions robotiques. La série Cosmos 3 avait précédemment lancé le 31 mai 2026 à la GTC Taipei le Cosmos 3 Nano (16 milliards de paramètres) et le Cosmos 3 Super (64 milliards de paramètres).","category":"Modèles","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA a publié Cosmos 3 Edge : un modèle de monde ouvert à 4 milliards de paramètres, capable d'inférer sur l'appareil et de générer des actions pour les robots - Aioga Actualités IA","description":"NVIDIA a lancé Cosmos 3 Edge, un modèle de monde ouvert de 4 milliards de paramètres, conçu pour fonctionner sur les appareils. Ce modèle peut aider les robots et les agents d'IA v...","url":"https://www.aioga.com/fr/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:23:12.835Z"},"de":{"title":"NVIDIA veröffentlicht Cosmos 3 Edge: Ein Open-World-Modell mit 4 Milliarden Parametern, das auf Geräten inferieren und Roboteraktionen generieren kann","summary":"NVIDIA hat Cosmos 3 Edge veröffentlicht, ein Open-World-Modell mit 4 Milliarden Parametern, das für den Einsatz auf Geräten entwickelt wurde. Dieses Modell kann Robotern und visuellen KI-Agenten helfen, die Umgebung zu verstehen, in Echtzeit zu schließen und Roboteraktionen lokal zu erzeugen. Die Cosmos-3-Serie hatte zuvor am 31. Mai 2026 auf der GTC Taipei Cosmos 3 Nano (16 Milliarden Parameter) und Cosmos 3 Super (64 Milliarden Parameter) vorgestellt.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA veröffentlicht Cosmos 3 Edge: Ein Open-World-Modell mit 4 Milliarden Parametern, das auf Geräten inferieren und Roboteraktionen generieren kann - Aioga KI-News","description":"NVIDIA hat Cosmos 3 Edge veröffentlicht, ein Open-World-Modell mit 4 Milliarden Parametern, das für den Einsatz auf Geräten entwickelt wurde. Dieses Modell kann Robotern und visuel...","url":"https://www.aioga.com/de/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:23:12.323Z"},"pt-BR":{"title":"NVIDIA lança Cosmos 3 Edge: modelo de mundo aberto com 4 bilhões de parâmetros, capaz de inferir e gerar ações de robôs no dispositivo","summary":"A NVIDIA lançou o Cosmos 3 Edge, um modelo de mundo aberto com 4 bilhões de parâmetros, projetado para execução em dispositivos. Este modelo pode ajudar robôs e agentes de IA visual a compreender o ambiente, realizar inferência em tempo real e gerar ações robóticas localmente. A série Cosmos 3 já havia lançado, em 31 de maio de 2026, no GTC Taipei, o Cosmos 3 Nano (16 bilhões de parâmetros) e o Cosmos 3 Super (64 bilhões de parâmetros).","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA lança Cosmos 3 Edge: modelo de mundo aberto com 4 bilhões de parâmetros, capaz de inferir e gerar ações de robôs no dispositivo - Aioga Notícias de IA","description":"A NVIDIA lançou o Cosmos 3 Edge, um modelo de mundo aberto com 4 bilhões de parâmetros, projetado para execução em dispositivos. Este modelo pode ajudar robôs e agentes de IA visua...","url":"https://www.aioga.com/pt-BR/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:23:51.005Z"},"ru":{"title":"NVIDIA выпустила Cosmos 3 Edge: открытая мировая модель с 4 миллиардами параметров, способная выполнять вывод и генерировать действия робота на устройстве","summary":"NVIDIA выпустила Cosmos 3 Edge, модель открытого мира с 4 миллиардами параметров, разработанную для работы на устройствах. Эта модель помогает роботам и визуальным ИИ-агентам понимать окружающую среду, осуществлять выводы в реальном времени и генерировать действия роботов локально. Серия Cosmos 3 ранее была представлена 31 мая 2026 года на GTC Taipei с моделями Cosmos 3 Nano (16 миллиардов параметров) и Cosmos 3 Super (64 миллиарда параметров).","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA выпустила Cosmos 3 Edge: открытая мировая модель с 4 миллиардами параметров, способная выполнять вывод и генерировать действия робота на устройстве - Aioga Новости ИИ","description":"NVIDIA выпустила Cosmos 3 Edge, модель открытого мира с 4 миллиардами параметров, разработанную для работы на устройствах. Эта модель помогает роботам и визуальным ИИ-агентам поним...","url":"https://www.aioga.com/ru/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:23:52.400Z"},"ar":{"title":"أصدرت NVIDIA نظام Cosmos 3 Edge: نموذج عالمي مفتوح يحتوي على 4 مليار معلمة، يمكنه الاستنتاج على الأجهزة وتوليد حركات الروبوت","summary":"أصدرت NVIDIA نموذج Cosmos 3 Edge، وهو نموذج عالم مفتوح يحتوي على 4 مليارات من المعاملات، مصمم خصيصًا للتشغيل على الأجهزة المحلية. يمكن لهذا النموذج أن يساعد الروبوتات ووكلاء الذكاء الاصطناعي البصري على فهم البيئة، وإجراء الاستدلال في الوقت الفعلي، وتوليد حركات الروبوت محليًا. وقد أطلقت سلسلة Cosmos 3 سابقًا نموذج Cosmos 3 Nano (16 مليار معاملة) وCosmos 3 Super (64 مليار معاملة) في 31 مايو 2026 خلال مؤتمر GTC Taipei.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"أصدرت NVIDIA نظام Cosmos 3 Edge: نموذج عالمي مفتوح يحتوي على 4 مليار معلمة، يمكنه الاستنتاج على الأجهزة وتوليد حركات الروبوت - Aioga أخبار الذكاء الاصطناعي","description":"أصدرت NVIDIA نموذج Cosmos 3 Edge، وهو نموذج عالم مفتوح يحتوي على 4 مليارات من المعاملات، مصمم خصيصًا للتشغيل على الأجهزة المحلية. يمكن لهذا النموذج أن يساعد الروبوتات ووكلاء الذكاء...","url":"https://www.aioga.com/ar/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:24:30.676Z"},"hi":{"title":"NVIDIA ने Cosmos 3 Edge लॉन्च किया: 40 अरब पैरामीटर वाला खुला दुनिया मॉडल, जो उपकरण पर अनुमान लगा सकता है और रोबोट क्रियाएँ उत्पन्न कर सकता है","summary":"NVIDIA ने Cosmos 3 Edge जारी किया, एक 4 अरब पैरामीटर वाला ओपन वर्ल्ड मॉडल, जिसे डिवाइस-एंड पर चलाने के लिए डिज़ाइन किया गया है। यह मॉडल रोबोट और विज़ुअल एआई एजेंट्स को पर्यावरण समझने, रीयल-टाइम तर्क करने और स्थानीय रूप से रोबोटिक क्रियाएं उत्पन्न करने में मदद कर सकता है। Cosmos 3 श्रृंखला को इससे पहले 31 मई 2026 को GTC Taipei में Cosmos 3 Nano (16 अरब पैरामीटर) और Cosmos 3 Super (64 अरब पैरामीटर) के साथ पेश किया गया था।","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA ने Cosmos 3 Edge लॉन्च किया: 40 अरब पैरामीटर वाला खुला दुनिया मॉडल, जो उपकरण पर अनुमान लगा सकता है और रोबोट क्रियाएँ उत्पन्न कर सकता है - Aioga AI समाचार","description":"NVIDIA ने Cosmos 3 Edge जारी किया, एक 4 अरब पैरामीटर वाला ओपन वर्ल्ड मॉडल, जिसे डिवाइस-एंड पर चलाने के लिए डिज़ाइन किया गया है। यह मॉडल रोबोट और विज़ुअल एआई एजेंट्स को पर्यावरण समझ...","url":"https://www.aioga.com/hi/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:24:36.104Z"},"it":{"title":"NVIDIA rilascia Cosmos 3 Edge: modello open world con 4 miliardi di parametri, capace di inferenza sul dispositivo e generazione di azioni dei robot","summary":"NVIDIA ha rilasciato Cosmos 3 Edge, un modello di mondo aperto con 4 miliardi di parametri, progettato per l'esecuzione su dispositivi. Questo modello può aiutare robot e agenti di intelligenza artificiale visiva a comprendere l'ambiente, fare inferenze in tempo reale e generare azioni robotiche localmente. La serie Cosmos 3 era stata precedentemente lanciata il 31 maggio 2026 alla GTC Taipei con Cosmos 3 Nano (16 miliardi di parametri) e Cosmos 3 Super (64 miliardi di parametri).","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA rilascia Cosmos 3 Edge: modello open world con 4 miliardi di parametri, capace di inferenza sul dispositivo e generazione di azioni dei robot - Aioga Notizie IA","description":"NVIDIA ha rilasciato Cosmos 3 Edge, un modello di mondo aperto con 4 miliardi di parametri, progettato per l'esecuzione su dispositivi. Questo modello può aiutare robot e agenti di...","url":"https://www.aioga.com/it/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:25:16.703Z"},"nl":{"title":"NVIDIA lanceert Cosmos 3 Edge: een open-wereldmodel met 4 miljard parameters, kan op apparaatniveau redeneren en robotacties genereren","summary":"NVIDIA heeft Cosmos 3 Edge uitgebracht, een openwereldmodel met 4 miljard parameters, speciaal ontworpen voor apparaatkantuitvoering. Dit model kan robots en visuele AI-agents helpen om de omgeving te begrijpen, realtime redeneringen uit te voeren en lokaal robotacties te genereren. De Cosmos 3-serie had eerder op 31 mei 2026 tijdens GTC Taipei Cosmos 3 Nano (16 miljard parameters) en Cosmos 3 Super (64 miljard parameters) geïntroduceerd.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA lanceert Cosmos 3 Edge: een open-wereldmodel met 4 miljard parameters, kan op apparaatniveau redeneren en robotacties genereren - Aioga AI-nieuws","description":"NVIDIA heeft Cosmos 3 Edge uitgebracht, een openwereldmodel met 4 miljard parameters, speciaal ontworpen voor apparaatkantuitvoering. Dit model kan robots en visuele AI-agents help...","url":"https://www.aioga.com/nl/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:25:12.143Z"},"tr":{"title":"NVIDIA, Cosmos 3 Edge'i yayınladı: 4 milyar parametreli açık dünya modeli, cihaz üzerinde çıkarım yapabiliyor ve robot hareketleri üretebiliyor","summary":"NVIDIA, cihazda çalışacak şekilde tasarlanmış 40 milyar parametreli açık dünya modeli Cosmos 3 Edge'i yayınladı. Bu model, robotların ve görsel AI ajanlarının ortamı anlamasına, gerçek zamanlı akıl yürütme yapmasına ve yerel olarak robot hareketleri üretmesine yardımcı olabilir. Cosmos 3 serisi daha önce 31 Mayıs 2026'da GTC Taipei'de Cosmos 3 Nano (16 milyar parametre) ve Cosmos 3 Super (64 milyar parametre) ile tanıtılmıştı.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA, Cosmos 3 Edge'i yayınladı: 4 milyar parametreli açık dünya modeli, cihaz üzerinde çıkarım yapabiliyor ve robot hareketleri üretebiliyor - Aioga AI Haberleri","description":"NVIDIA, cihazda çalışacak şekilde tasarlanmış 40 milyar parametreli açık dünya modeli Cosmos 3 Edge'i yayınladı. Bu model, robotların ve görsel AI ajanlarının ortamı anlamasına, ge...","url":"https://www.aioga.com/tr/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:25:59.017Z"},"vi":{"title":"NVIDIA ra mắt Cosmos 3 Edge: mô hình thế giới mở 4 tỷ tham số, có thể suy luận trên thiết bị và tạo hành động cho robot","summary":"NVIDIA ra mắt Cosmos 3 Edge, một mô hình thế giới mở với 4 tỷ tham số, được thiết kế để chạy trên thiết bị. Mô hình này có thể giúp robot và các thực thể AI thị giác hiểu môi trường, suy luận theo thời gian thực và tạo ra các hành động của robot tại chỗ. Dòng Cosmos 3 trước đó đã ra mắt Cosmos 3 Nano (16 tỷ tham số) và Cosmos 3 Super (64 tỷ tham số) tại GTC Taipei vào ngày 31 tháng 5 năm 2026.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA ra mắt Cosmos 3 Edge: mô hình thế giới mở 4 tỷ tham số, có thể suy luận trên thiết bị và tạo hành động cho robot - Tin tức AI Aioga","description":"NVIDIA ra mắt Cosmos 3 Edge, một mô hình thế giới mở với 4 tỷ tham số, được thiết kế để chạy trên thiết bị. Mô hình này có thể giúp robot và các thực thể AI thị giác hiểu môi trườn...","url":"https://www.aioga.com/vi/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:25:59.777Z"},"id":{"title":"NVIDIA merilis Cosmos 3 Edge: model dunia terbuka dengan 4 miliar parameter, dapat melakukan inferensi di perangkat dan menghasilkan tindakan robot","summary":"NVIDIA merilis Cosmos 3 Edge, sebuah model dunia terbuka dengan 4 miliar parameter, yang dirancang untuk dijalankan di perangkat. Model ini dapat membantu robot dan agen AI visual memahami lingkungan, melakukan penalaran secara real-time, dan menghasilkan aksi robot secara lokal. Seri Cosmos 3 sebelumnya telah meluncurkan Cosmos 3 Nano (16 miliar parameter) dan Cosmos 3 Super (64 miliar parameter) pada 31 Mei 2026 di GTC Taipei.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA merilis Cosmos 3 Edge: model dunia terbuka dengan 4 miliar parameter, dapat melakukan inferensi di perangkat dan menghasilkan tindakan robot - Berita AI Aioga","description":"NVIDIA merilis Cosmos 3 Edge, sebuah model dunia terbuka dengan 4 miliar parameter, yang dirancang untuk dijalankan di perangkat. Model ini dapat membantu robot dan agen AI visual...","url":"https://www.aioga.com/id/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:26:36.770Z"},"th":{"title":"NVIDIA เปิดตัว Cosmos 3 Edge: โมเดลโลกเปิดที่มีพารามิเตอร์ 4 พันล้านตัว สามารถอนุมานและสร้างการเคลื่อนไหวของหุ่นยนต์บนอุปกรณ์","summary":"NVIDIA เปิดตัว Cosmos 3 Edge ซึ่งเป็นโมเดลโลกเปิดที่มีพารามิเตอร์ 4 พันล้านตัว ออกแบบมาเพื่อใช้งานบนอุปกรณ์ โมเดลนี้สามารถช่วยให้หุ่นยนต์และตัวแทน AI ทางสายตาเข้าใจสภาพแวดล้อม ทำการวิเคราะห์แบบเรียลไทม์ และสร้างการเคลื่อนไหวของหุ่นยนต์ในเครื่องได้ ซีรีส์ Cosmos 3 เคยเปิดตัว Cosmos 3 Nano (พารามิเตอร์ 16 พันล้านตัว) และ Cosmos 3 Super (พารามิเตอร์ 64 พันล้านตัว) ไปแล้วเมื่อวันที่ 31 พฤษภาคม 2026 ในงาน GTC Taipei","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA เปิดตัว Cosmos 3 Edge: โมเดลโลกเปิดที่มีพารามิเตอร์ 4 พันล้านตัว สามารถอนุมานและสร้างการเคลื่อนไหวของหุ่นยนต์บนอุปกรณ์ - ข่าว AI Aioga","description":"NVIDIA เปิดตัว Cosmos 3 Edge ซึ่งเป็นโมเดลโลกเปิดที่มีพารามิเตอร์ 4 พันล้านตัว ออกแบบมาเพื่อใช้งานบนอุปกรณ์ โมเดลนี้สามารถช่วยให้หุ่นยนต์และตัวแทน AI ทางสายตาเข้าใจสภาพแวดล้อม ทำกา...","url":"https://www.aioga.com/th/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:26:46.315Z"},"pl":{"title":"NVIDIA wprowadza Cosmos 3 Edge: model otwartego świata z 4 miliardami parametrów, który może wykonywać wnioski na urządzeniu i generować akcje robota","summary":"NVIDIA ogłosiła Cosmos 3 Edge, model świata otwartego o 4 miliardach parametrów, zaprojektowany do działania na urządzeniach końcowych. Model ten może pomagać robotom i inteligentnym agentom wizualnym AI w rozumieniu środowiska, dokonywaniu wnioskowania w czasie rzeczywistym oraz lokalnym generowaniu działań robota. Seria Cosmos 3 została wcześniej przedstawiona 31 maja 2026 roku na GTC Taipei, prezentując Cosmos 3 Nano (16 miliardów parametrów) i Cosmos 3 Super (64 miliardy parametrów).","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA wprowadza Cosmos 3 Edge: model otwartego świata z 4 miliardami parametrów, który może wykonywać wnioski na urządzeniu i generować akcje robota - Aioga Wiadomości AI","description":"NVIDIA ogłosiła Cosmos 3 Edge, model świata otwartego o 4 miliardach parametrów, zaprojektowany do działania na urządzeniach końcowych. Model ten może pomagać robotom i inteligentn...","url":"https://www.aioga.com/pl/news/cmrucw21u0xgubi7fmmu88bby/","contentTranslated":true,"sourceHash":"d20973db96a470fe","translatedAt":"2026-07-23T01:27:31.881Z"}}}}