Gasgoo Munich-BeingBeyond, a developer of general embodied foundation models, has signed a cooperation agreement with PaXini, officially becoming a partner in the PaXini Inside ecosystem.
Under the agreement, the two will combine BeingBeyond’s leading algorithms in general embodied foundation models with PaXini’s core strengths in high-precision, multi-dimensional tactile perception and massive multimodal data. The partnership will focus on comprehensive collaboration across model training, real-world proof-of-concept (POC) verification, and algorithm optimization.
Image Source: PaXini
Prior to this, the two companies had already forged deep operational ties.
On July 28, BeingBeyond unveiled Being-H0.8, an implicit tactile world action model. Pre-trained on more than 500,000 hours of human video data, it marks the first integration of tactile modalities into large-scale model pre-training—giving robots a more complete understanding and predictive capability for physical interactions.
During this development, PaXini served as a critical data partner, supplying the high-quality multimodal data essential to the effort.
Beyond technology and operations, the pair are jointly pushing for industry standards, aiming to define the future of embodied intelligence starting at the data source.
According to PaXini, robots entering the physical world require a closed loop spanning perception, data, models, and applications: sensing the world, generating data through interaction, building capabilities via models, and continuously iterating in real-world scenarios. Around this loop, PaXini has established a full-stack layout covering multi-dimensional tactile sensors, embodied data collection and processing, robot product development, and scenario applications.
Leveraging its market-leading multi-dimensional tactile sensors and a "human-centric" multimodal perception collection system, PaXini has built OmniSharingDB—a multimodal embodied intelligence dataset covering vision, touch, force, and movement trajectories.
PaXini has also built a network of large-scale data factories across North, East, Central, Southwest, and South China. This infrastructure supports a complete pipeline—from perception hardware and data collection to cleaning, labeling, quality verification, and large-scale delivery—generating nearly 10 billion multimodal data points with rapid turnaround capabilities.
Building on this foundation, PaXini officially launched the "Global Embodied Ecology Alliance & PaXini Inside Partner Plan" on the eve of the 2026 World Robot Congress. Targeting innovators, robotics firms, research institutes, scenario partners, channel partners, and industrial capital, the initiative involves a substantial investment of hundreds of millions in yuan and resources. By opening up technology, capital synergy, scenarios, markets, and industry linkages, PaXini aims to drive these technologies toward mass adoption.
Several companies across the supply chain have already joined the initiative.
Since its founding in May 2025, BeingBeyond has hit several key milestones centered on "large-scale human video pre-training." It has released the world’s first embodied foundation models pre-trained on thousands, tens of thousands, 200,000, and 500,000 hours of human video, respectively. The company has also established a full-stack infrastructure spanning data pipelines, pre-training, post-training, evaluation, and edge deployment.
