NVIDIA · Driving VLA model

NVIDIA Alpamayo 2 Edge and AlpaJudge

NVIDIA Alpamayo 2 Edge is a 2B-parameter adaptive reasoning vision-language-action model for autonomous driving, built on the Nemotron-based backbone of Cosmos 3 Edge with a diffusion action expert. It generates Chain-of-Causation reasoning only when needed and runs at 160-190 ms pixel-to-trajectory latency on NVIDIA DRIVE AGX Thor with TensorRT optimisation. Closed-loop reinforcement learning in AlpaGym improves driving performance by 48.9% over the best checkpoint without it. The companion AlpaJudge model provides human-calibrated evaluation and reward scoring of reasoning traces, and both models are published on Hugging Face.

Open the source ↗

Similar Agents & Automation releases

Learning Heterogeneous Tactile Representation for General Dexterous Manipulation
Bing Han, Yichuan Peng, Haoyu Dong, Yuanmeng Hu, Yongyi Jia, Fangjing Wang, Kecheng Zheng, Yiming Zhang (Robbyant) · Tactile Foundation Model · Week 47, 2026
World In-Context Learning for Agentic Robot Manipulation
Heng Fang, Zehui Guo, Yuehan Zhang, Mingyang Du, Dingkang Liang, Xiang Bai (H-EmbodVis) · In-Context Robot Learning · Week 47, 2026
Gemini Robotics 2.5
Google DeepMind · VLA Robotics Model · Week 47, 2026
Toward the Era of Agentic Embodied Intelligence
Chao Liang, Ke Chen, Fengxiaoxiao Li, Shibo Huang, Yufeng Sun, Zhaokai Yin, Jianhuan Lai, Haozhe Feng (AutoLab) · Agentic Robot Control · Week 47, 2026
GR00T 2
NVIDIA · Open World Action Model · Week 47, 2026
FLEX
MIT and New York University researchers (led by Andreea Bobu, MIT) · LLM Robot Planning · Week 47, 2026

NVIDIA Alpamayo 2 Edge and AlpaJudge

Similar Agents & Automation releases

More from NVIDIA