X Square Quanta X1 Universal Wheeled Dual Arm Humanoid Robot

X Square Quanta X1: The Wheeled Dual-Arm Robot That Completed the First Fully Autonomous AI-Powered Outdoor Food Delivery

The Quanta X1 is X Square's first-generation robot platform, designed for "precision manipulation in service environments" according to Pandaily's April 2026 interview with WANG Qian, and powered by the company's WALL-A vision-language-action (VLA) foundation model.

The Quanta X1 achieved the most commercially significant milestone in X Square's documented public history on or before January 2026: a fully autonomous outdoor food delivery mission completed without any human intervention, where the robot "successfully navigated indoor and outdoor tasks to complete a delivery in an open environment" while handling "challenges such as strong winds, deformed packaging, and visual occlusions," confirmed by X Square's Series A++ funding announcement documented across multiple independent sources including The Robot Report, PRNewswire, and Interesting Engineering.

Technically, the X1 features a wheeled chassis with lightweight robotic arms providing a working range of up to 1 meter, a maximum operating speed of 2 meters per second, and 20 degrees of freedom. The WALL-A model that powers it combines VLA systems with world models to form an integrated architecture that enables zero-shot generalization: the ability to perform tasks the robot has not been specifically trained on, demonstrated during the logistics parcels mission where the robot identified irregular items it had not encountered in training.

X Square Robot: The Company

Founded 2023, Shenzhen-Based, Full-Stack AI and Hardware

X Square Robot was founded in 2023 in Shenzhen, China, and positions itself as a "pioneer in general purpose embodied AI." The company operates under the stated thesis, confirmed by The Robot Report's September 2025 coverage, that "robots capable of reliably performing unpredictable real-world tasks, such as those in homes or hotels, are far from commercially viable" due to two specific barriers: "an over-reliance on task-specific training data and a disproportionate focus on bipedal locomotion."

The Quanta X1's wheeled chassis design directly reflects this second critique: by choosing a wheeled base over bipedal legs, X Square invested its engineering effort in the manipulation and AI capabilities that enable real-world task generalization rather than in the bipedal locomotion challenge that consumes significant engineering resources at competing companies for a mobility capability that most indoor service environments do not require.

The Autonomous Food Delivery Milestone

What Made It Significant

Multiple independent sources documenting X Square's January 2026 Series A++ announcement describe the same milestone in consistent terms: "a landmark breakthrough in autonomous food delivery, powered by Wall-A, where the Quanta X1 successfully navigated indoor and outdoor tasks to complete a delivery in an open environment."

The specific challenges the Quanta X1 encountered and overcame, each representing a different category of real-world complication:

Strong winds: Outdoor environment dynamic perturbation that affects physical robot stability and may affect object handling. A robot that cannot maintain task completion during wind events is not deployable in actual outdoor service environments.

Deformed packaging: The packaging of a real delivery order may not match the pristine training data condition of standard boxes. Deformed packaging requires the robot to reason about the object's current physical state rather than relying on learned patterns from ideal-condition training.

Visual occlusions: Objects partially hidden from the robot's cameras require inference about what cannot be directly perceived. X Square's description confirms: "the robot used the model's causal inference to 'fill in the blanks' when objects were partially hidden," demonstrating a world model capability that extends the robot's perception beyond what direct sensor observation provides.

Autonomous self-correction: "When encountering operational stalls or friction, the robot autonomously self-corrects and completes the task loop without any human intervention." This indicates the robot can detect task execution failures and recover from them, rather than halting and requiring human restart.

This combination of outdoor environment navigation, physical manipulation in degraded conditions, causal inference under occlusion, and autonomous self-correction represents a qualitatively higher bar than controlled indoor demonstrations.

Zero-Shot Generalization in Logistics

The second documented Quanta X1 capability demonstration is in complex logistics: "facing piles of parcels, the robot utilizes zero-shot generalization to identify irregular items." Zero-shot generalization means handling items the robot has not seen during training, recognizing their properties from the WALL-A model's general understanding rather than from specific training examples.

For logistics deployment, zero-shot generalization is the capability that separates commercially viable logistics robots from laboratory demonstrations: real logistics environments involve the full diversity of package types, sizes, shapes, and conditions that cannot be fully enumerated in a training dataset.

Technical Specifications

Confirmed Specifications

From Interesting Engineering's January 2026 coverage (citing X Square's official specifications) and Pandaily's April 2026 interview:

Type: Wheeled bimanual robot (dual-arm)

Chassis: Wheeled (not bipedal)

Degrees of Freedom: 20 DoF total

 Arm Working Range: Up to 1 meter

 Maximum Speed: 2 m/s

AI Model: WALL-A (closed full-stack) and WALL-OSS (open-source version) 

Dexterous Hands: Configured with ArtiXon 20 DoF hands (in integrated system)

Applications: Service environments, food delivery, logistics, household tasks

WALL-A: VLA with World Models

The WALL-A model, which powers the Quanta X1, is described by Interesting Engineering as a "vision-language-action (VLA) model that combines VLA systems with world models to form an integrated architecture." The specific technical innovation of WALL-A over standard VLA models is the world model integration: "By using world models to predict actions and causal reasoning to interpret feedback, the system improves a robot's ability to generalize to new tasks without prior training."

Standard VLA models directly map visual observations and language instructions to actions based on training data. WALL-A adds a world model that predicts the future state of the environment given a proposed action, enabling the robot to evaluate planned actions before executing them and to reason about unobserved parts of the environment through causal inference, the mechanism that enabled the Quanta X1 to "fill in the blanks" when objects were visually occluded.

WALL-OSS: Open-Source Democratization

In September 2025, X Square released WALL-OSS, described as "an open-source version of its model family designed to democratize embodied intelligence and accelerate community-driven innovation." WALL-OSS has been integrated into Hugging Face's LeRobot framework, enabling developers using the LeRobot ecosystem to access X Square's model architecture alongside the broader LeRobot model community.

The open-source release strategy reflects X Square's data flywheel approach: community adoption of WALL-OSS generates diverse deployment scenarios and feedback that informs the improvement of the closed WALL-A model, creating the self-reinforcing data and model evolution cycle that WANG Qian described as essential for the next phase of embodied AI competition.

Data Pipeline and Training Infrastructure

"First in China to Scale Real-World Data Resources"

X Square Robot claims to be "the first company in China to scale up real-world data resources" for robotics, having developed "advanced data capture tools including teleoperation, exoskeletons, and Universal Manipulation Interface (UMI)." These three data capture modalities serve different purposes:

Teleoperation: A human operator controls the robot remotely, with the robot's sensor data and actions recorded as training demonstrations.

Exoskeletons: Human operators wearing exoskeletons perform manipulation tasks, with the exoskeleton recording the human's arm and hand movements as training data that can be mapped to the robot's kinematics.

Universal Manipulation Interface (UMI): A data capture tool developed and open-sourced by Stanford that enables high-quality manipulation demonstrations without requiring the robot to be present during data collection.

The combination of these three capture modalities enables X Square to generate manipulation training data across a broader range of task types and environments than any single capture modality alone, supporting the WALL-A model's cross-environment generalization capability.

Applications and Deployment Verticals

Autonomous food delivery: The January 2026 autonomous outdoor delivery mission.

Logistics: The zero-shot parcel identification demonstration.

Senior healthcare: Named as a deployment vertical without specific documentation of a comparable demonstration milestone.

Advanced manufacturing: Named as a deployment vertical, consistent with the Quanta X2's specification for household and industrial environments.

Quanta X1 vs. Quanta X2

The Quanta X1 and Quanta X2 are X Square's two robot platforms, with the X2 being the successor platform launched in August 2025:

Quanta X1: Wheeled bimanual robot, 20 DoF, 1-meter arm reach, 2 m/s speed, lighter configuration, designed for service environments and precision manipulation.

Quanta X2: Wheeled humanoid, 62 DoF, 1.64 meters tall, 25 kg dual-arm payload, full humanoid form factor for household and industrial environments.

The X1 is the established commercial deployment platform with documented real-world mission completions. The X2 is the next-generation platform for the broader range of environments that require humanoid-scale physical capability.

Summary

The X Square Quanta X1 is the wheeled dual-arm bimanual robot that became the documented platform for the first fully autonomous AI-foundation-model-powered outdoor food delivery mission, powered by the WALL-A VLA with world model architecture and backed by $240 million in total funding from ByteDance, HongShan, Alibaba, and Meituan. Its 20 DoF, 1-meter arm reach, 2 m/s speed, and demonstrated zero-shot generalization across irregular logistics parcels collectively establish the Quanta X1 as the most operationally validated wheeled bimanual service robot from China's embodied AI startup sector, with WALL-OSS's Hugging Face LeRobot integration making X Square's underlying AI architecture available to the global robotics research and development community.