NVIDIA Releases Open 3D Vision Models for Robots

On September 15, 2026, NVIDIA released two open 3D perception models on Hugging Face — FoundationStereo for stereo depth and FoundationPose for 6-DoF object pose, both under the NVIDIA Open Model License with commercial use rights and operating zero-shot without fine-tuning.

Compiled by Sergey KostenchukPublished 2026-09-16Updated 2026-09-18

2026-09-16 Research NVIDIA

🤖 NVIDIA Releases Open 3D Vision Models for Robots

On September 15, NVIDIA released two open 3D perception models on Hugging Face: FoundationStereo (63 million parameters) for stereo depth and FoundationPose for 6-DoF object pose. Both are under a license with commercial application rights and were highlighted at CVPR.

🌍 Research 3D perception models have become ready-made ONNX and TensorRT weights with a license for commerce. Robotics teams don't need datasets and fine-tuning: zero-shot inference of depth and pose works from Jetson Orin to H100.

👤 With a stereo camera and NVIDIA hardware, the weights can be downloaded for free from Hugging Face: FoundationStereo requires a pair of frames, FoundationPose — a CAD model in OBJ and an RGBD frame. Transparent and shiny objects, overexposure, and poor lighting are weak points.

Source 1: https://huggingface.co/nvidia/c-foundationstereo-s Source 2: https://huggingface.co/nvidia/foundationpose

Sources