SFLARE × Zongheng Physics: Full-Stack Embodied Intelligence Partnership
The partners are building an end-to-end collaboration across data pipelines, VAM action models, robot hardware, and large-scale store deployment.
In May in Beijing, SFLARE and Zongheng Physics (Beijing) Technology Co., Ltd. signed a full-stack strategic partnership for embodied intelligence, connecting data pipelines, VAM video-to-action models, robot hardware, and terminal deployment into a complete human data → model training → robot execution → ten-thousand-store rollout loop.
Why This Partnership
The hardest part of deploying embodied intelligence is not perfecting one link, but making data, models, hardware, and scenarios work well together.
Zongheng Physics has deep full-stack expertise. Its VAM video-to-action model supports pretraining, post-training, interactive real-time trajectories, and unlimited streaming generation, closing the loop from data to models to real-machine deployment.
SFLARE focuses on consumer and commercial robot platforms. Together with the Institute of Automation, Chinese Academy of Sciences, we have built autonomous decision-making from environmental perception through intent understanding to task execution, while developing mature controllers, actuators, sensors, interaction systems, scaled manufacturing, and growing consumer and commercial channels.
The strengths are complementary: Zongheng Physics supplies data and working models; SFLARE supplies producible hardware and deployable scenarios.
What We Will Build Together
Dexterous-Manipulation Data Pipeline
AI glasses will collect data continuously in real stores while robot hardware supports physical data collection and annotation, creating a dexterous-manipulation dataset for VAM training.
VAM Video-to-Action Model
The partners will train a data-native VAM model and deploy it directly to SFLARE controllers, closing the video input → action generation → physical execution loop.
Scaled Deployment across More Than Ten Thousand Stores
Zongheng Physics will open access to more than ten thousand stores for priority SFLARE deployments, reference scenarios, and repeatable commercial delivery.
Embodied Intelligent Robots
The companies will connect home scenarios, appliance ecosystems, and whole-home intelligence, combining dedicated VAM models with mass-production capability.
Human–Machine Interaction and Consumer Ecosystem
They will jointly develop voice, vision, touch, and intent-understanding interactions across homes, hotels, eldercare, and communities.
Bringing embodied intelligence into every home has always been our goal. This partnership accelerates both the technical foundation and scenario deployment.
SFLARE robots are already operating in smart eldercare, providing health management, fully autonomous indoor navigation and obstacle avoidance, and remote video control. The ecosystem is expanding from homes to communities and professional institutions.
The single-arm Xinghu H series focuses on family health and institutional service; the dual-arm Xingzhi S series targets research data collection and advanced demonstrations. Both support three-finger grippers or five-finger dexterous hands and align closely with Zongheng Physics' data and model system.
Signing is only the beginning. Together the companies have built an integrated data–model–hardware–scenario advantage and will pursue:
- More capable VAM action models
- More human and real-machine data
- More cost-effective robot hardware
- Faster deployment in stores and homes
Our aim is embodied intelligence that is genuinely useful, easy to use, and affordable.