On April 30, Beijing ShengShu Technology Co., Ltd. (hereinafter referred to as “ShengShu Technology”) and Astribot (Shenzhen) Co., Ltd. (hereinafter referred to as “Astribot”) announced a strategic partnership. The two parties will conduct long-term collaborative innovation around embodied intelligence and world model directions, accelerating AI from “understanding the world” to “acting in the world.”
This partnership focuses on three key directions:
First, building an intelligent collaboration system centered on the parties’ proprietary general world models and embodied models. The two parties will collaborate on layered coordination of cognition and action, driving embodied models to jointly complete the “perception – decision-making – execution” closed loop, enhancing robots’ understanding, reasoning, and generalization capabilities in complex, non-standard scenarios.
Second, building a real-world data-driven model evolution mechanism. Leveraging ShengShu Technology’s world model capabilities in processing massive heterogeneous multimodal data, combined with Astribot’s scaled deployment capabilities in commercial services, home, and industrial scenarios, the two parties will continuously accumulate high-quality embodied interaction and physical trajectory data, forming a real-world data flywheel to drive continuous generalization and capability leaps of the embodied brain in dynamic environments.
Third, advancing vertical scenario deployment and commercial applications. The two parties will jointly explore application paths for embodied intelligence in retail, service, home, and other scenarios, accelerating the scaled deployment of cutting-edge technologies in real-world environments and promoting the industrialization of general physical intelligence agents.
Yesterday, ShengShu Technology officially released its general-purpose brain for the physical world — the World Action Model Motubrain. As the industry’s first model to unify “predicting the world” and “executing actions” into a single modeling framework, Motubrain can not only “see” but also “do,” breaking through the data wall limitations of traditional embodied models. Currently, Motubrain has demonstrated four core disruptive capabilities: “one brain, multiple abilities” (handling massive complex tasks), “one brain, multiple forms” (cross-platform adaptation for various robots), “one brain, end-to-end” (completing ultra-long-horizon tasks in a single inference), and “one brain, foresight” (dynamic decision-making and automatic error correction), marking that AI is crossing the generation stage to become an actor in the real physical world.
In this partnership and frontier model exploration, Astribot, as the world’s only mass-produced tendon-driven AI robot, provided critical support for Motubrain’s R&D. Thanks to its unique tendon-drive design mimicking human muscle-tendon transmission, its kinematic structure and optimized motion control characteristics are highly isomorphic with humans, naturally aligning action distributions with human demonstration data and reducing the distribution shift and adaptation cost from human data to physical execution for world models. Unlike traditional rigid motors, the rigid-flexible coupling of tendon drives gives the S1 flexible safety, dexterous manipulation, and agile response capabilities, demonstrating exceptional “hand-following” performance in actual teleoperation. This not only ensures high stability and human-like expressiveness in real environments and complex interactions, but also provides world models with unique high-quality, zero-latency continuous flexible physical interaction data, serving as a key bridge connecting high-level semantics with low-level physical control.
Since its founding, Astribot has upheld a “Design for AI” integrated hardware-software system architecture and a “top-tier hardware – leading teleoperation (data) – efficient models” full-stack proprietary R&D approach. The core of this architecture lies in deeply coupling “AI intelligence” with “best-in-class manipulation,” committed to enabling robots to learn, think, and act like humans. In 2025, the company successively launched its proprietary strong-reasoning foundation model Lumo-1 and the fast-slow system model DuoCore deployed in real retail scenarios nationwide, building a technology system centered on embodied models. On this foundation, Astribot continues to expand model collaboration capabilities, opening compatibility and coordination with leading multimodal world models to further improve task execution efficiency and stability in complex environments.
Going forward, the two parties will build a commercialization and data dual-loop system around “model – data – hardware – scenarios.” Starting from “small-loop validation,” they will progressively advance to joint R&D and specialized collaboration, continuously driving the fusion and deployment of embodied models and world models in the real physical world, and converting technical value into industrial value.
About ShengShu Technology
Founded in 2023, ShengShu Technology is the world’s first company to achieve unification of the digital and physical worlds through general world models. With its foundation world model at the core, in the digital world, it builds the multimodal large model Vidu based on the World Generation Model, continuously improving multi-temporal generation and interaction capabilities and driving the intelligent evolution of content production; in the physical world, it has launched the World Action Model Motubrain, connecting the perception, prediction, and action pipeline to accelerate the development of embodied intelligence toward general intelligent agents.
About Astribot
The definer of tendon-driven AI robots and the world’s first company to mass-produce tendon-driven AI robots. Since its founding in 2022, Astribot has upheld a “Design for AI” integrated hardware-software system architecture and a “top-tier hardware – leading teleoperation (data) – efficient models” full-stack proprietary R&D approach. The company’s unique tendon-drive design mimics human muscle-tendon motion and force control, giving robots high dexterous manipulation capability, high human-like expressiveness, and high interaction safety — particularly suited for complex dexterous operations and close human-interaction scenarios. Currently, the robots have been deployed across research, commercial services, entertainment performances, and industrial sectors, combining AI large models with specific scenario exploration to accelerate robot industry applications and commercialization.