Skip to content
NewA calmer, faster ZeroHour
StoryModel releaseinfo

NVIDIA presents GR00T 2 at CoRL with disputed success rates

2 sources · 2 articlesFirst reported Updated ()1
What's new

First merged summary.

Merged summary · x-ai/grok-4.7 · rewritten as coverage arrivesSummaries by Vani AI

NVIDIA unveiled GR00T 2 at CoRL, but the two reports disagree on success rates and training-data volume.

At CoRL in Austin, NVIDIA introduced GR00T 2, a Cosmos-based world-action model that, according to The Decoder, predicts the next movement and a short future camera view rather than mapping images and language directly to motors as GR00T 1 did. NVIDIA reports 71 percent average success across 50 tasks with one demonstration and 95 percent with five. The Decoder reports rates from 10 to 80 percent, a rise from 40 to 95 percent after under three minutes of training, and at least 97 percent on toy assembly with a planner. Hours of human video also differ: NVIDIA cites 100,000 hours of open egocentric data, while The Decoder cites about 400,000 hours and more than a thousand GB300 chips. NVIDIA also highlighted Isaac Lab 3.0, EgoHand, and partners Siemens and Foxconn.

Key facts
  • GR00T 2 completed 40 of 50 tasks at least once without task-specific demonstrations.
  • NVIDIA ranks GR00T 2 top three on RoboCasa365, RoboLab, RoboDojo, and MolmoSpaces.
  • Isaac Lab 3.0 with Newton reached up to 2.5x Isaac Lab 2.3 with PhysX throughput.
  • EgoHand had over 40,000 Hugging Face downloads in two weeks after launch.
  • Without a language model, toy-car assembly succeeded in 23 of 100 attempts.
  • GRAIL and RoboTTT are two of six CoRL Best Paper Award finalists.
Explain this AI news with Vani AIWhat was announced, what is new and why it matters, in a new chat

What each outlet adds

beyond the shared facts
  • NVIDIA BlogCoRL rankings, Isaac Lab speedup, EgoHand downloads, and paper finalists.
  • The DecoderGR00T 1 contrast, GB300 training, touch limits, and Jetson Thor wipe rate.
Vendors
NVIDIANvidia
Products
Isaac LabNewtonNVIDIA CosmosJetson Thor
Organizations
SiemensFoxconnNexuniMicroAGIOpenAI
AI models
GR00T 2EgoHandGR00T 2 TurboGPT-6 Astra

Coverage

oldest first · one row per article · UTC
  1. · 1d ago NVIDIA BlogFirst report

    NVIDIA introduces GR00T 2 robot world action model at CoRL

    NVIDIA says GR00T 2 hit 71% average success across 50 tasks with one demonstration and 95% with five.

  2. · 6h ago The Decoder

    Nvidia's GR00T 2 turns human video into robot actions

    Lab success rates ranged from 10 to 80 percent; with a planner, toy assembly reached at least 97 percent of trials.