What comes in the box?

The robot, the API and SDK, and SimReady URDF and USD assets. End effectors are ordered separately.

Can I fit my own end effector?

Yes. We publish the mechanical and electrical interface, so a third-party or custom tool can be easily integrated.

Why wheels instead of legs?

Safety, stability, speed, and runtime. A wheeled base can't fall over, stays planted for two-handed work, moves faster across a floor, and spends no battery on balancing, which is how Vega 1 Pro runs a full shift.

Why not a cheaper DIY rig?

DIY is fine for a first experiment. Vega is built to be relied on, with the build quality, runtime, and support to match.

Does Vega 1 Pro come with AI?

No. It ships as hardware with Python, C++, and Rust APIs and the SDK. The task behavior is yours to build or teach, and Vega comes with the tools to collect demonstration data that is ready for physical AI model training.

Which labs have published research on Vega 1 Pro?

Teams at UC Berkeley, Stanford, USC, Carnegie Mellon, UIUC, Northwestern, Purdue, Tsinghua, Michigan, NVIDIA, Google DeepMind, Toyota Research Institute, Samsung SDS Research America, Microsoft Research, and Amazon FAR have published work built on Vega. Fourteen entries so far, with summaries, authors, and links, are collected under Research Papers.

General-Purpose Wheeled Humanoid

Meet Vega 1 Pro

Collect. Teleop. Simulate. Deploy. From the lab bench to the factory floor.

Order Now Download Spec Sheet

Features

One Robot for the Whole Job

Teleoperate over VR or exoskeleton, capture demonstrations, and simulate with SimReady URDF and USD. One pip install, and you're running on day one.

An engineer teleoperating Vega 1 Pro's hands at a bench

Capabilities

Explore What It Can Do

Bimanual Dexterity

Two 7-DoF arms, each carrying 10+ lbs (5 kg) in any pose, so Vega 1 Pro handles two-handed tasks a single-arm rig can't.

Specifications

The Full Spec

Type: General-purpose wheeled humanoid
Reach: Floor to 7 ft 2 in (2.2 m)
Degrees of Freedom: 24
Arms14 (7 each)
Torso3
Head3
Base4 (2 steering, 2 drive)
Runtime: 20 hrs of operation
See the benchmark video
Weight: 298 lbs (135 kg)
Dimensions: Folded (L × W × H)26 × 23 × 26 in (0.66 × 0.56 × 0.66 m)
Max height (arms raised)8 ft 2.6 in (2.50 m)
Compute: NVIDIA Jetson AGX Thor
Interfaces: USB2× USB-A, 2× USB-C
Network1× 5GbE Ethernet
Display1× DisplayPort
Port details in the hardware setup guide on DexHub.
Software: Python, C++, and Rust APIs
SimReady models (URDF/USD)
Fleet management
VR and exoskeleton teleoperation
SDK docs on DexHub. URDF and USD models on GitHub →.

Arms: Two arms, 7 DoF each (14 total)
Payload: 10+ lbs (5 kg) per arm
Torso: 3 DoF, adjustable
End Effectors: Ordered separately
DexGripper S or D
Your own tool
Tool Mounting: Published mechanical and electrical interface for third-party tools
Arm Sensing: Position-controlled
Optional force/torque (F/T) add-on for admittance control

Head: 3 DoF
Head Sensors: Stereo RGB camera
Mobile Base: Omni-directional
Base Sensors: 2× LiDAR, one front and one rear
4× RGB cameras

Download Spec Sheet

In Practice

What Teams Build with Vega 1 Pro

The workflows teams run on Vega today, from collecting demonstrations to testing in simulation.

Collect Bimanual Demonstration Data

Capture demonstration data for imitation learning, teleoperated over VR or exoskeleton, in the space the skill will run.

PGDG: Data from One Demonstration

Who It's For

From the Research Bench to the Factory Floor

Research Labs
Teams looking for an easy-to-use dual-arm mobile manipulator.
Frontier AI Labs
Model teams working at the cutting edge of physical AI.
Product Teams
Teams moving fast to build and deploy physical AI solutions.
Industrial Applications
Application teams bringing safe, robust, and reliable robotic solutions to industry.

Built for Developers

The Support You Can Trust

The model, the skill, and the application are yours to build. Vega 1 Pro supplies what sits underneath: open interfaces, an open community, and hardware you can count on.