Facts

• Company: NVIDIA
• Product type: Software
• Status: development
• Website: https://www.nvidia.com/en-us/ai-data-science/robotics/

Profile sections

• Summary
• Overview
• Market context
• Audience
• Workflow
• Capability
• Deployment environment
• Positioning tags
• Specifications
• Record / Articles
• Timeline / Research
• Sources

Summary

Sonic is NVIDIA generalist humanoid controller research project for physical AI. It trains policies from large motion-capture datasets and deploys onboard humanoid robots, supporting whole-body motion, human-motion imitation and real-time robot control experiments.

Overview

Sonic is NVIDIA generalist humanoid controller research project for physical AI. It trains policies from large motion-capture datasets and deploys onboard humanoid robots, supporting whole-body motion, human-motion imitation and real-time robot control experiments.

Market context

Sonic is NVIDIA's robot control and policy model, focused on real-time decision-making and action execution.

• Target environment: Robotics research environments, humanoid simulation workflows and NVIDIA robotics developer ecosystems.
• Workflow context: Whole-body control, humanoid motion policy development, controller testing and training-to-deployment robotics workflows.
• Customer context: Humanoid developers, robotics researchers and teams building on NVIDIA's robotics software stack.
• Deployment model: Development-stage software layer associated with NVIDIA humanoid robotics research.
• Commercial maturity: Research and development product with public positioning, not a standalone commercial robot.
• Adoption constraints: Adoption depends on developer access, hardware compatibility, simulation fidelity, latency and real-robot validation.
• Market position: Control-policy layer within NVIDIA's broader robotics stack, adjacent to GR00T, Isaac and Cosmos rather than robot hardware.
• Adjacent products: Project GR00T, NVIDIA Isaac, Google RT-series models, Figure Helix

Audience

Developers

Workflow

Whole-body control, humanoid motion policy development, controller testing and training-to-deployment robotics workflows.

Capability

Software

Deployment environment

Robotics research environments, humanoid simulation workflows and NVIDIA robotics developer ecosystems.

Positioning tags

• Audience: Developers
• Workflow: Whole-body control, humanoid motion policy development, controller testing and training-to-deployment robotics workflows.
• Capability: Software
• Deployment environment: Robotics research environments, humanoid simulation workflows and NVIDIA robotics developer ecosystems.

Specifications

• Core Functions: Real-time control policy execution; Low-latency action decision systems; Integration with simulation-trained robot models
• Platform Type: software
• Deployment Model: research
• Platform Role: humanoid_control_layer
• System Type: humanoid_control_framework
• System Control: Real-time control policies
• System Deployment: Action execution layer
• System Integration: Integration with simulation-trained models
• System Capabilities: Low-latency decision systems

Record / Articles

• 2026-06-22 - Article - The Humanoid Bottleneck Is Leaving the Cage

Timeline / Research

• 2026-02-21 - Market Signal - NVIDIA Open-Sources SONIC for Humanoid Whole-Body Control - SONIC is a generalist humanoid controller trained on over 100 million frames of diverse motion data.
• 2026-02-19 - Research Publication - NVIDIA releases GEAR-SONIC whole-body control stack for humanoid robots - NVIDIA released GEAR-SONIC through the GR00T Whole-Body Control repository as a generalist humanoid whole-bod…