Facts

• Company: Sudo AI
• Product type: Robotics Foundation Model
• Status: research_preview
• Capability focus: Simulation Trained Manipulation Policy
• Website: https://www.sudo.ai/

Profile sections

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

Summary

sudo R1 is Sudo AI humanoid robot platform for embodied-AI research and applied automation. It is positioned for manipulation, mobility and real-world task learning where developers need a human-scale robot body for testing AI control systems.

Overview

sudo R1 is Sudo AI humanoid robot platform for embodied-AI research and applied automation. It is positioned for manipulation, mobility and real-world task learning where developers need a human-scale robot body for testing AI control systems.

Market context

sudo R1 is a simulation-trained manipulation model system focused on object picking.

• Target environment: Manipulation benches, logistics-style picking tasks and robot-learning research environments.
• Workflow context: Closed-loop object picking, sim-to-real transfer, clutter handling and reactive grasp execution.
• Customer context: Robotics teams studying contact-rich manipulation and simulation-first robot policy development.
• Deployment model: Company technical reveal rather than a broad commercial software platform.
• Commercial maturity: Early public demonstration with detailed reported evaluation but limited customer evidence.
• Adoption constraints: Adoption depends on hardware availability, reproducibility, task coverage and transfer beyond picking.
• Market position: Manipulation-focused robot foundation-model lane, adjacent to Helix and GENE-26.5.
• Adjacent products: GENE-26.5, Helix, Project GR00T

Audience

Developers

Workflow

Closed-loop object picking, sim-to-real transfer, clutter handling and reactive grasp execution.

Capability

Simulation-trained manipulation policy

Deployment environment

Manipulation benches, logistics-style picking tasks and robot-learning research environments.

Positioning tags

• Audience: Developers
• Workflow: Closed-loop object picking, sim-to-real transfer, clutter handling and reactive grasp execution.
• Capability: Simulation-trained manipulation policy
• Deployment environment: Manipulation benches, logistics-style picking tasks and robot-learning research environments.

Specifications

• Control Model: Simulation-to-real manipulation control
• Model Type: manipulation-centric foundation model system
• Task Scope: object picking across diverse real-world object conditions
• Training Data Source: simulation data only; no real-world demonstrations
• Closed Loop Control: 15 to 25 Hz
• First Attempt Success Rate: ~98 percent in Sudo AI reported testing
• Evaluation Window: 60 minutes
• Software Stack: Simulation-trained manipulation model deployed to physical robot hardware
• Supported Tasks: Object picking across unseen rigid, deformable, transparent, reflective and irregular objects
• Deployment Model: Robot manipulation system trained without per-object tuning or real-world demonstration data

Tags

• Peer Group: Robot foundation model

Record / Articles

• 2026-04-20 - Article - Sudo AI reveals sudo R1, a manipulation system trained entirely in simulation

Timeline / Research

• 2026-04-20 - Capability - Sudo AI publicly reveals sudo R1 simulation-trained manipulation system - Sudo AI introduced sudo R1 as a fully integrated robot system for object picking, built with self-developed hardware an…