X Square
X Square Robot is Shenzhen's foundation-model powerhouse — a $2.8 billion company backed by ByteDance, Alibaba, Meituan, and HongShan, building the WALL family of embodied AI models (including China's leading open-source WALL-OSS) and the Quanta wheeled robots now bookable by real households for cleaning services. This guide covers the X Square lineup, technology, applications, and what buyers should know before purchasing.
X Square Robot
Introduction / Overview
X Square Robot (Shenzhen) is what happens when a robotics company decides the robot is the easy part. Founder-CEO Wang Qian's thesis — "the next phase of competition in embodied intelligence is essentially a battle of foundation models built on data closed-loops" — has attracted the most concentrated strategic backing in Chinese embodied AI: four consecutive funding rounds to a $2.8 billion valuation, with Meituan leading the Series A, Alibaba Cloud and HongShan co-leading a ~$100M A+, and ByteDance and HongShan anchoring the $140 million (RMB 1 billion) A++ closed in January 2026 — plus a Forbes China AI Top 50 listing. The product of that capital is the WALL model family, the field's most complete embodied-AI stack: WALL-A, the end-to-end foundation model whose shared-attention, task-routed architecture trains on real robot data plus augmented generative video; WALL-B (April 2026), built on a World Unified Model architecture training perception, language, action, and physical prediction in one network rather than modular VLA parts; WALL-WM, the event-level world model; and WALL-OSS — China's leading open-source embodied AI model, integrated into Hugging Face's LeRobot framework, with WALL-OSS-0.5 achieving over 80% autonomous completion on real-robot tasks without any post-training.
The hardware carries the models into the world. The Quanta X1 — a wheeled dual-arm robot for precision service manipulation — delivered food autonomously across indoor and outdoor environments, handling strong winds, deformed packaging, and visual occlusions, using causal inference to "fill in the blanks" behind occluded objects and self-correcting through stalls without human intervention, then zero-shot-identifying irregular parcels in logistics piles. The Quanta X2 (August 2025) is the flagship "robot butler": 172 cm with 756 mm arm reach and 6 kg end-effector payload on a 10B+ parameter learning stack, sized for household furniture, hotel workspaces, and light industrial stations. The data engine beneath is China's first at scale: teleoperation, exoskeletons, and Universal Manipulation Interface (UMI) capture tools feeding a model-driven pipeline. And commercialization is unusually literal: in 2026 X Square launched a bookable home-cleaning robot service with 58.home — consumers ordering embodied AI into their actual houses — alongside deployments across manufacturing, logistics, eldercare, hotels, retail, and public services, and the convening power of EAIDC 2026, the world's first large-scale embodied-AI developers conference, hosted in Shenzhen. For buyers and developers alike, X Square is the foundation-model-first bet on general-purpose robots.
Design and Features
The WALL Model Family — The Real Product
- WALL-A — the end-to-end embodied brain: a shared-attention + task-routed feed-forward architecture that observes, analyzes, and channels information to specialized pathways simultaneously — trained on real robotic action data plus augmented generative video for robustness across lighting, object shapes, and environments, with fine motor control in hands and upper body as an explicit strength.
- WALL-B — the World Unified Model generation: perception, language, action, and physical prediction trained in one unified network, yielding stronger multimodal understanding, spatial reasoning, and continual learning from real-world interactions — the architectural break from modular VLA.
- WALL-WM — event-level world modeling: aligning language, vision, and action around meaningful events for cross-modal learning and physical prediction.
- WALL-OSS — the open-source democratizer on GitHub and Hugging Face (LeRobot-integrated), generalizing across robot types and solving catastrophic forgetting and vision-language-action sync — with the 0.5 release clearing 80%+ autonomy on real tasks untuned.
Quanta X2 — The Butler Flagship
- Human-scale service anatomy — 172 cm, 756 mm reach, 6 kg end payload: dimensioned for countertops, hotel carts, shelving, and light industrial stations.
- Force-controlled bimanual manipulation — precision dual-arm work for long-horizon household and service tasks on the 10B+ parameter stack.
- Wheeled by conviction — X Square's stated position: the industry over-focuses on bipedal locomotion when generalized manipulation is the actual barrier — the X2's wheels buy deployment reliability where it counts.
Quanta X1 — The Proven Operator
The wheeled dual-arm precision platform with the field's most instructive demo record: open-environment autonomous delivery through wind, packaging deformation, and occlusion; causal-inference reasoning; autonomous self-correction through friction and stalls; and zero-shot generalization across irregular logistics parcels.
The Data Flywheel
China's first scaled real-world data operation: teleoperation rigs, exoskeleton capture, and UMI tools generating manipulation data, amplified by a model-driven pipeline producing high-quality training data at scale — the closed loop of hardware, data, and models Wang identifies as the race's decisive weapon.
The Ecosystem Play
WALL-OSS open-sourcing plus EAIDC — the world's first dedicated embodied-AI developers conference — positioning X Square's models as the platform the community builds on, every integration feeding the flywheel.
Technology and Specifications
The Stack at a Glance
| Layer | Products | Signature |
|---|---|---|
| Foundation models | WALL-A, WALL-B, WALL-WM | World Unified Model; continual learning |
| Open source | WALL-OSS / OSS-0.5 | 80%+ untuned real-task autonomy; LeRobot |
| Hardware | Quanta X1, Quanta X2 | 172 cm / 756 mm / 6 kg butler class |
| Data | Teleop, exoskeletons, UMI | China's first scaled real-world corpus |
Confirm current datasheets before specifying — models and hardware iterate quarterly, and datasheets are available on request.
Why Unified Beats Modular
Modular VLA stacks translate between separately trained parts, leaking capability at every seam; WALL-B's unified training makes physical prediction native to the same network that perceives and acts — the architecture behind the X1's occlusion reasoning and self-correction, and the credible path to robots that improve from deployment rather than despite it.
Position in the Market
Among your wheeled service humanoids, X Square stakes the foundation-model pole: Galbot owns retail deployment scale, UniX AI the household-hardware first — X Square owns the model layer itself, with the only open-source stack of the three, consumer-bookable services live, and a valuation that says the strategic giants agree.
Applications and Use Cases
Household Services
The 58.home partnership making X Square the rare humanoid company consumers can book: home cleaning delivered by embodied AI, with the butler roadmap beyond it.
Logistics and Delivery
X1-class autonomous delivery and zero-shot parcel handling across the open-environment conditions that defeat scripted systems.
Manufacturing and Industry
Precision bimanual manipulation at machine-tending, kitting, and light-station scale.
Eldercare, Hotels, Retail, and Public Services
The declared commercialization spread, riding the models' cross-scenario generalization.
Developer and Research Programs
WALL-OSS on Hugging Face, the LeRobot integration, and EAIDC's convened community: the most accessible serious embodied-AI stack available.
Advantages / Benefits
- The model advantage — the WALL family's unified architecture and China's leading open-source embodied AI.
- Proof under weather — outdoor autonomous delivery with occlusion reasoning and self-correction: generalization demonstrated, not claimed.
- Consumer-real commercialization — bookable household service, the industry's rarest credential.
- The data moat — China's first scaled real-world capture operation feeding a closed-loop flywheel.
- Strategic-giant conviction — ByteDance, Alibaba, Meituan, and HongShan across four rounds to $2.8B.
- The ecosystem chair — open source plus the world's first embodied-AI developers conference.
For buyers researching where to buy X Square robots, or comparing Quanta X1 and X2 prices and costs against Galbot and other wheeled humanoids, the evaluation should weigh model-layer capability, generalization evidence, open-source access, data-flywheel depth, and commercialization reality — the dimensions where X Square's foundation-model strategy leads.
FAQ
What is X Square Robot?
X Square Robot is a Shenzhen embodied-AI company valued at $2.8 billion — backed by ByteDance, Alibaba, Meituan, and HongShan across four rounds — building the WALL family of embodied foundation models (WALL-A, WALL-B, WALL-WM, and the open-source WALL-OSS) and the Quanta X1 and X2 wheeled robots.
What is the Quanta X2?
The Quanta X2 is X Square's flagship wheeled humanoid "robot butler": 172 cm tall with 756 mm arm reach and 6 kg end-effector payload on a 10B+ parameter learning stack — force-controlled bimanual manipulation sized for homes, hotels, and light industrial stations.
What is WALL-OSS?
WALL-OSS is X Square's open-source embodied foundation model — among China's leading — available on GitHub and Hugging Face and integrated into the LeRobot framework; the WALL-OSS-0.5 release achieved over 80% autonomous completion on real-robot tasks without any post-training.
What is WALL-B?
WALL-B (April 2026) is X Square's World Unified Model foundation: training perception, language, action, and physical prediction within a single unified network rather than modular VLA components — enabling stronger spatial reasoning and continual learning from real-world interactions, complemented by the WALL-WM event-level world model.
Can consumers actually use X Square robots?
Yes — through the 2026 partnership with 58.home, X Square's home-cleaning robot service is bookable by real households, bringing embodied AI into consumer homes as a purchasable service alongside enterprise deployments.
What has the Quanta X1 demonstrated?
Autonomous food delivery across indoor and outdoor environments — handling strong winds, deformed packaging, and visual occlusions via causal inference, self-correcting through operational stalls without human help — plus zero-shot identification of irregular parcels in logistics piles.
Why is X Square Robot important?
X Square is executing the purest foundation-model strategy in robotics: a unified model architecture, China's largest real-world data flywheel, the leading open-source stack, and consumer-live commercialization — the bet, backed to $2.8 billion, that models win the embodied race.
How much do X Square robots cost?
The Quanta platforms occupy the professional wheeled-humanoid tier, with WALL-OSS freely available to developers and services priced per engagement. Contact a distributor for current pricing, deployment guidance, datasheets, and delivery timelines for your region.
Summary
X Square Robot bet everything on the layer you can't see: WALL models that treat perception, language, action, and physics as one problem, trained on China's first scaled real-world data operation and proven by a robot that delivered food through wind and occlusion, reasoning past what it couldn't see and correcting what went wrong — no human in the loop. With the Quanta X2 butler standing 172 centimeters over its 6-kilogram grip, WALL-OSS open to every developer on Hugging Face, households booking cleaning robots through 58.home, and ByteDance-to-Alibaba capital compounding to $2.8 billion, X Square is embodied AI's foundation-model thesis in full execution. For anyone looking to buy a Quanta X1 or X2, build on WALL-OSS, compare their costs against Galbot and other wheeled humanoids, or plan a household, logistics, or industrial deployment, this guide provides the essential foundation.