GR00T N1 2B

Training compute

7.1×10²² FLOP

Parameters

2.2B

Published

Mar 18, 2025

GR00T N1 2B is an AI model developed by NVIDIA (United States), first published in March 2025. It works in the robotics, vision and language domain, on tasks such as robotic manipulation and animal (human/non-human) imitation.

Training it took an estimated 7.1×10²² FLOP of compute (estimation method: hardware). The model has 2,190,000,000 parameters. Training ran on 1,024 NVIDIA H100 SXM5 80GB for about 48.83 hours.

Access: Open weights (non-commercial). Its weights are openly available. It is built on top of Eagle 2. The reference paper has 738 citations. Epoch AI rates the confidence of this record as confident.

Full record

Organization: NVIDIA
Country of organization: United States
Domain: Robotics, Vision, Language
Task: Robotic manipulation, Animal (human/non-human) imitation
Training compute: 7.1×10²² FLOP
Compute estimation method: Hardware
Parameters: 2,190,000,000
Training hardware: NVIDIA H100 SXM5 80GB
Chips used: 1,024
Training time: 49 h
Training power draw: 1.4 MW
Numerical format: BF16
Model accessibility: Open weights (non-commercial)
Open weights: Yes
Base model: Eagle 2
Citations: 738
Epoch confidence: Confident