X Square Robot Launches QUANXTA Zero Series to Build Integrated Embodied Data Production System
X Square Robot has launched the QUANXTA Zero Series, an integrated embodied AI platform that unifies data collection, cleaning, annotation, training and evaluation to address key bottlenecks in robotics data scale, quality and efficiency. The system includes three hardware configurations and an end-to-end software pipeline, with the flagship G1 enabling high-efficiency, synchronized data capture of up to 100 demonstrations per hour and direct conversion of raw data into model training workflows.
The system integrates data collection, cleaning, annotation, training and evaluation into a single embodied AI production pipeline.
The company says the platform aims to overcome persistent bottlenecks in robotics data scale, quality and efficiency.
X Square Robot has launched its QUANXTA Zero Series, a software-hardware integrated platform designed to streamline embodied AI data production, as the robotics industry continues to grapple with persistent data bottlenecks that limit model training and deployment.
The company said the new system aims to address structural challenges in embodied intelligence development, where data scarcity, low collection efficiency and inconsistent quality remain key constraints despite rapid advances in foundation models.
QUANXTA Zero is positioned as a full-stack workflow that connects data collection, synchronization, annotation, training and evaluation into a unified production loop.
The launch comes as embodied AI developers increasingly shift focus from model architecture toward scalable, high-quality data pipelines, which are widely viewed as critical to achieving general-purpose robotic intelligence.
At the center of the system is a new product matrix designed for multi-scenario data capture. The QUANXTA Zero Series includes three configurations: the G1, a flagship UMI-VIO system combining a headband with dual grippers for balanced performance and endurance; the G0, a VR-based mobile system using a PICO 4 headset, dual-gripper setup and backpack for full-body teleoperation and precise localization; and the E0, a lightweight ego-centric capture device equipped with six cameras for first-person data collection.
The company said the G1 system enables rapid collection of embodiment-free manipulation data across locomotion, dual-arm control and object handling tasks, capturing visual, tactile and audio signals in a synchronized framework.
It supports multi-view perception, combining first-person and wrist-level perspectives to preserve fine-grained interaction details such as grasping, contact dynamics and occlusion.
According to X Square Robot, the system achieves sensor synchronization accuracy within one millisecond and frame-level alignment across video streams, enabling high-fidelity replay across different robotic embodiments.
In internal testing, the system reportedly supports close to 100 demonstrations per hour, representing a significant efficiency improvement over conventional teleoperation setups.
The hardware is designed for continuous operation, featuring a lightweight configuration that eliminates backpack systems and relies on a headband-plus-gripper architecture. A magnetic hot-swappable battery system enables uninterrupted field deployment.
Beyond hardware, QUANXTA Zero integrates a full software pipeline intended to convert raw demonstrations into structured training data. The company said its app layer provides a task marketplace and operational dashboard that allows users to initiate, monitor and manage data collection without manually decomposing workflows.
The downstream data pipeline includes multi-source synchronization with microsecond-level timestamp alignment, automated cleaning and multimodal annotation that filters low-value segments such as pauses and failed attempts, and a hybrid AI-human quality control system that routes uncertain samples for manual review while automating high-confidence labeling. The system also includes privacy and security features such as facial blurring, watermarking and access controls.
Finally, processed data feeds directly into model training and evaluation workflows, enabling iterative performance feedback loops that identify weak scenarios and guide targeted data collection.
X Square Robot said the platform is intended to shift embodied data production from fragmented hardware-based collection toward an industrialized, end-to-end data engine that supports scalable robot learning and general-purpose embodied intelligence development.
Source
1 X Square Robot Newsroom. "X Square Robot Launches QUANXTA Zero Series, Reframing Embodied Data Production Through Universal Manipulation Interfaces" July 2, 2026. https://x2robot.com/en/news/6a46341cc7feadddbc603a33
2 品玩 《自变量发布QUANXTA Zero系列无本体数采方案，“数采元年”定义数采基建，可降低数采成本60%》，July 1, https://www.pingwest.com/a/315228
