DexForce open-sources EmbodiChain for embodied-model training with fully generated data
Evidence notes
- DexForce announced the open-sourcing of EmbodiChain, its embodied-model training framework built around fully generated data.
- The company positioned the release as part of its Engine-driven Sim2Real VLA approach, aimed at reducing dependence on scarce real-world robot data.
- The move extended DexForce's embodied-intelligence stack beyond robot hardware into developer and training infrastructure for model creation.
Company context
DexForce builds wheeled humanoid robots and physical-AI systems for industrial and commercial service tasks. The Shenzhen company raised a RMB1B Series B in June 2026 at a post-money valuation above RMB10B, with deployment activity across manufacturing and service environments.