DigestAI news desk

Cut through the AI noise.

Generative AI & Models4 min read

NVIDIA releases Metatron v1 unified humanoid control policy for Unitree G1

NVIDIA has made Metatron v1 available, a unified ONNX‑based control policy that handles walking, scene interaction and recovery for the Unitree G1 humanoid robot. The model package includes a 459,455,983‑byte checkpoint, configuration files and separate browser and racer exports, all licensed under the NVIDIA Open Model License with source code under Apache‑2.0. The release is aimed at…

1 source primary source

Key points

  • NVIDIA released Metatron v1, a unified ONNX policy for walking, scene interaction, and recovery on the Unitree G1 robot.
  • The checkpoint is 459,455,983 bytes and is licensed under the NVIDIA Open Model License; source code remains Apache‑2.0.
  • NVIDIA warns the exports are not safety‑certified, not a collision‑avoidance system, and require independent validation for real‑robot use.

The distribution notes stress that the exported files are not interchangeable, are not safety‑certified, and should not be treated as a collision‑avoidance or complete training‑reproduction package. Users must validate hardware limits, provide emergency‑stop mechanisms and avoid feeding raw demo inputs directly to the network. NVIDIA recommends using MuJoCo WASM for simulation and ONNX Runtime Web for inference, with WebGPU preferred for real‑time operation. The model’s numerical export checks show a maximum absolute action error of 5.55e-6 across 60 test states, but no real‑robot benchmarks are provided.

Full story from huggingface.co · via Reddit AI communities primary sourceOpen source ↗

Metatron v1: NVIDIA's Unified Humanoid Control Policy for the Unitree G1

huggingface.co · 27 October 2026

Loading the full article…

This text was published by huggingface.co. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗

Coverage and discussion

1source
Topics · follow one to build your own front page

The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.

Comments

via GitHub Discussions

More in Generative AI & Models

All →

Related stories