Gasgoo Munich-BeingBeyond, a developer of general embodied foundation models, has officially entered a strategic partnership with Daimon Robotics.
The alliance merges their complementary strengths in modeling, perception, and data. Together, they will collaborate on multimodal data development, model training, and scenario validation — aiming to accelerate the deployment of robotics in precision manufacturing and flexible production.
Specifically, their capabilities align precisely: BeingBeyond’s foundation model powers high-level task understanding and planning, while Daimon-TWM’s tactile anchoring world model sharpens real-time reaction and error correction. Combined, the system allows robots to not only "comprehend" tasks but also sense changes and adjust movements instantly — boosting the adaptability, stability, and precision of complex operations.
On the implementation front, the partners will prioritize proof-of-concept (POC) validation in precision manufacturing and flexible production. Targeting high-precision tasks like flexible object handling, precision assembly, and adaptive plugging, they aim to bridge the gap from capability verification to large-scale replication — building a general intelligence foundation that works across different robot bodies and workstations.
Image Credit: BeingBeyond
BeingBeyond was the first in China to propose using large-scale human video data to train general embodied model frameworks, and it is also the country’s first AI startup to launch a native implicit world action model.
The company has already released the Being-H general dexterous manipulation model and the Being-M general mobile manipulation model, pushing pre-training volume to 500,000 hours. It has built a full-stack infrastructure spanning data pipelines, pre- and post-training, evaluation, and edge deployment, while establishing ecosystem partnerships with dozens of industry players, including Lingxinqiaoshou, Zhipingfang, and DiGu Robotics.
Daimon Robotics focuses on physical interaction perception. Its flagship product, Daimon-TWM, is the world’s first "physical interaction brain" designed to drive dexterous manipulation. By unifying physical cognition, decision-making, and instant control within a single framework, the system can predict contact evolution and failure risks — enabling millisecond-level corrections before errors occur.
The industrial deployment of embodied intelligence currently faces a significant capability gap. Pure vision solutions struggle with the precision demands of contact-based tasks, while single-point tactile solutions lack high-level task understanding — leaving many high-precision workstations without viable automation alternatives.
By moving beyond the traditional "single-company full-stack" approach, the two firms are stitching together capabilities through an ecosystem model of "general brain plus specialized tactile sensing." This strategy is expected to lower the technical barriers for real-world deployment.
