{"rewrite":{"id":"r_17383630c4eb337e5500818c","clusterId":"c_4b9f4e0d27f75e885fe2bc44","slug":"nvidia-announces-gr00t-2-robot-model-that-learns-tasks-from-five-demonstrations","model":"deepseek-v4-1-flash","headline":"NVIDIA Announces GR00T 2 Robot Model That Learns Tasks From Five Demonstrations","summary":"NVIDIA announced GR00T 2, a world action model for operating robots, trained on 402,730 hours of video of human movement, 10,184 hours of handheld gripper data, and 23,500 hours of robot teleoperation data. NVIDIA says a new task can be learned from five human demonstrations and executed in 30 minutes on a GeForce RTX 5090, which can also run inference. Related code is scheduled for release in December 2026.","whyItMatters":"NVIDIA is pitching the training data volume and the five-demonstration, 30-minute learning loop as the path from demonstration footage to working robot tasks, with the code release set for December 2026.","webCardHtml":"\u003cp\u003eNVIDIA\u0026#39;s GR00T 2 is trained on three separate data sets: 402,730 hours of video of human movement, 10,184 hours of handheld gripper operation, and 23,500 hours of human teleoperation of robots. The model handles whole-body motion, moving beyond hand movement alone.\u003c/p\u003e\u003cp\u003eNVIDIA describes tasks including preparing breakfast, assembling a PC, and lighting a match. In one demonstration the robot retrieves a cup from the floor, wipes spilled liquid with a towel underfoot, places the cup in a sink, and drops the towel in a collection box. GR00T 2 can also coordinate with other AI models, as in a Tower of Hanoi example run with GPT-6 Astra.\u003c/p\u003e","blueskyPost":"GR00T 2 learns a task from five demonstrations and runs it in 30 minutes on a GeForce RTX 5090, so the training data is the product. NVIDIA trained it on 402,730 hours of human movement video.","twitterPost":"GR00T 2 runs inference on a GeForce RTX 5090, which puts robot task learning on a desktop card rather than a datacenter.","threadsPost":"GR00T 2's 30-minute task learning runs on a GeForce RTX 5090, the same card a person can buy. NVIDIA trained the model on 402,730 hours of human movement video and 10,184 hours of handheld gripper data, and the code arrives in December 2026.","newsletterBlurb":"NVIDIA announced GR00T 2, a world action model for robots trained on 402,730 hours of human movement video, 10,184 hours of handheld gripper data, and 23,500 hours of robot teleoperation data. NVIDIA says a new task can be learned from five human demonstrations and run in 30 minutes on a GeForce RTX 5090. Related code is scheduled for December 2026.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20261117-nvidia-robot-gr00t-2/\",\"title\":\"NVIDIA Announces \\\"GR00T 2,\\\" an AI That Teaches Humanoid Robots Human Movement, Learning Tasks From a Few Demonstrations\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":5659,"outputTokens":787,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1794908426,"createdAt":"2026-11-17T09:35:58.000Z","publishedAt":"2026-11-17T09:39:19.000Z","updatedAt":"2026-11-17T09:39:19.000Z"},"cluster":{"id":"c_4b9f4e0d27f75e885fe2bc44","canonicalTitle":"人型ロボットに人間の動きを学習させるためのAI「GR00T 2」をNVIDIAが発表、数回のお手本からタスクを学べる","representativeArticleId":"a_701259b7862c529ad1b5773c","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[\"GR00T 2\"],\"studios\":[\"NVIDIA\"],\"people\":[],\"type\":\"announcement\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-11-17T08:54:00.000Z","lastSeenAt":"2026-11-17T08:54:00.000Z","updatedAt":"2026-11-17T09:39:14.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20261117-nvidia-robot-gr00t-2/","title":"人型ロボットに人間の動きを学習させるためのAI「GR00T 2」をNVIDIAが発表、数回のお手本からタスクを学べる"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":["GR00T 2"],"studios":["NVIDIA"],"people":[],"type":"announcement","domain":"other","is_roundup":false},"keyFacts":["NVIDIA announced GR00T 2, a world action model for operating robots.","GR00T 2 was pretrained on 402,730 hours of video recording human movements.","NVIDIA says a new task can be learned from five human demonstrations and executed in 30 minutes on a GeForce RTX 5090.","Related code for GR00T 2 is scheduled to be released in December 2026.","Part of the training data is published on Hugging Face as nvidia/PhysicalAI-Human-1M-100K."],"pageLinks":{"healthSlugs":["gr00t-2","nvidia"],"storySlugs":[],"entities":[{"slug":"gr00t-2","supported":0,"disputed":0,"contradicted":0,"unverified":3,"overall":"unverified","story_count":1,"has_dossier":false},{"slug":"nvidia","supported":6,"disputed":0,"contradicted":1,"unverified":38,"overall":"red","story_count":54,"has_dossier":true}],"stories":[],"entitiesOk":true,"storiesOk":true}}
