X Square Robot and AGIBOT unveil unified embodied AI models for home

X Square Robot announced its new embodied AI foundation model, Wall‑B, designed to power home robots within weeks, and AGIBOT declared 2026 as the first year of large‑scale commercial deployment of its unified physical intelligence architecture【1】. Both firms are positioning unified models that combine vision, language, action and physics prediction as a way to move robots from controlled labs into real‑world settings.

Key takeaways

• X Square’s Wall‑B uses a World Unified Model that trains vision, language, action and physical prediction jointly, mimicking infant learning【1】.
• The company plans to place home robots in everyday households within 35 days of the announcement【1】.
• AGIBOT’s “One Robotic Body with Three Intelligences” framework integrates motion, interaction and operation intelligence into a single system for scalable deployment【2】.
• AGIBOT marks 2026 as “Deployment Year One,” highlighting the rollout of its 10,000th robot in March 2026【2】.
• Both firms cite a data‑driven flywheel—development‑state for X Square and deployment‑state for AGIBOT—to accelerate improvement and productivity【1】【2】.

X Square Robot’s World Unified Model for Home Robots

At a Beijing launch titled “Born to Bot, Bot to Family,” X Square Robot introduced Wall‑B, the first full implementation of its World Unified Model (WUM) architecture【1】. Unlike modular approaches that train perception, language and control separately, WUM optimizes these capabilities within a single network from day one, allowing the model to predict physical outcomes such as force and friction before acting【1】. The company says the model was trained on data from real, non‑staged home environments to expose it to the “long tail” of household scenarios, from misplaced objects to spontaneous human activity【1】. Demonstrations included a robot arranging flowers while adjusting its grip in real time, a task completed without pre‑set trajectories【1】. X Square acknowledges that current systems still require remote intervention for mistakes, but it expects continuous operation and data collection to drive rapid improvement【1】.

AGIBOT’s Unified Physical Intelligence Architecture

In Shanghai, AGIBOT declared 2026 as the inaugural year of large‑scale commercial deployment of embodied AI, emphasizing a shift from proving robot capabilities to delivering measurable productivity at scale【2】. Its “One Robotic Body with Three Intelligences” framework unifies motion, interaction and operation intelligence within a single robotic platform, enabling perception, decision‑making and execution to operate under real‑world constraints such as force, precision and safety【2】. AGIBOT presented a full‑series portfolio covering humanoids, wheeled platforms and multi‑form robots, all built on this unified architecture【2】. The company highlighted a “deployment‑state data flywheel” that will accelerate value creation as robots are rolled out across industries, noting the rollout of its 10,000th robot in March 2026 as evidence of scaling capability【2】.