DexForce
About
About DexForce
DexForce develops wheeled humanoid robots and physical-AI software for manufacturing and commercial service work. Its W1 Pro combines an omnidirectional mobile base with an upper body and two arms, giving the platform human-like reach without the balance and energy demands of bipedal walking.
The robot supports teleoperation and autonomous task development for handling, inspection and machine-side workflows. Cameras, onboard compute and manipulation controls form a reusable platform that can move between indoor work areas. DexForce develops the hardware, control stack and application software while advancing deployments with industrial partners.
Articles
Products
Funding
- 2026-06-30 - Series B - DexForce closes RMB1 billion Series B
valuation: > CNY 10B (~$1.47B) - The financing valued DexForce above RMB10 billion after the investment. - source: dexforce.com
- 2025-07-10 - A1 and A2 rounds - DexForce announces A1 and A2 financing rounds
The two completed rounds together raised several hundred million yuan for embodied-intelligence development. - source: dexforce.com
- 2024-05-14 - Strategic investment - DexForce completes strategic financing led by Lenovo Capital
The investment supported product development, team expansion and market growth for DexForce’s embodied-intelligence business. - source: 10100.com
Investors
- Lenovo Capital - Lead - VC - 2 funding events - repeat backer - first seen 2024-05-14 - latest 2025-07-10
- Chengdu Sci-tech Innovation Investment - Participant - Unknown - 1 funding event - first seen 2026-06-30
- Chengdu Science & Technology Venture Capital - Lead - VC - 1 funding event - first seen 2025-07-10
- Guiyang Digital Economy Fund - Participant - VC - 1 funding event - first seen 2026-06-30
- Hengjian Asset - Participant - Angel / Individual - 1 funding event - first seen 2026-06-30
- Hongtai Fund - Lead - VC - 1 funding event - first seen 2025-07-10
- ICBC Capital - Participant - VC - 1 funding event - first seen 2026-06-30
- Lens Technology - Participant - Angel / Individual - 1 funding event - first seen 2026-06-30
- Nanshan Emerging Industry Fund - Participant - VC - 1 funding event - first seen 2025-07-10
- Nanshan Strategic Emerging Industry Investment - Participant - Unknown - 1 funding event - first seen 2026-06-30
- Qianhai Mother Fund - Participant - VC - 1 funding event - first seen 2026-06-30
- Shenzhen Capital Group - Participant - VC - 1 funding event - first seen 2026-06-30
- Sichuan Academician Fund - Participant - VC - 1 funding event - first seen 2026-06-30
- Sichuan Science and Technology Investment Group - Participant - Unknown - 1 funding event - first seen 2025-07-10
- Tanyuan Venture Capital - Participant - VC - 1 funding event - first seen 2025-07-10
- Tianying Capital - Participant - VC - 1 funding event - first seen 2025-07-10
- Yicun Songling - Participant - Angel / Individual - 1 funding event - first seen 2025-07-10
- Zhurui Capital - Participant - VC - 1 funding event - first seen 2026-06-30
Team
- Jia Kui - Founder - His spatial-intelligence and Sim2Real research background underpins DexForce’s physical-AI and humanoid robot stack.
Research
- DexForce coauthors UNITAS 3D-native world action model
2026-10-08 - Research Publication - The paper reports 99.8% LIBERO, 89.1% LIBERO-Plus and 92.94% / 93.04% RoboTwin Clean / Randomized success. Its real-robot Table 2 reports 18/20 book-organizing and 16/20 ball-packing trials, averaging 85% over these two tasks. The public project includes interactive scene predictions; checkpoints, inference and training code remain planned releases.
Source - DexForce open-sources EmbodiChain robot-training framework
2026-01-20 - Research Publication - Real2Sim extracts physical priors from limited real examples, while Gen2Sim builds three-dimensional scenes and assets from language instructions. The pipeline combines visual augmentation, randomized physics and automatic recovery trajectories after failed grasps; code and technical documentation were released publicly.
Source
Timeline
Events
- 2026-10-08 - Research Publication - DexForce coauthors UNITAS 3D-native world action model
The paper reports 99.8% LIBERO, 89.1% LIBERO-Plus and 92.94% / 93.04% RoboTwin Clean / Randomized success. Its real-robot Table 2 reports 18/20 book-organizing and 16/20 ball-packing trials, averaging 85% over these two tasks. The public project includes interactive scene predictions; checkpoints, inference and training code remain planned releases.
Source - 2026-09-22 - Partnership - DexForce partners with Bosuda on new-energy manufacturing robots
The planned loop extracts production tasks, reconstructs them in simulation, trains models and returns robots to factory testing. Robot operating data would feed subsequent training; DexForce separately described feeding, package flipping, scanning and packing work at other SMT sites.
Source - 2026-09-18 - Partnership - DexForce partners with Today International on industrial logistics
Planned joint work covers research, scenario validation and engineering for picking, transport, palletising and machine loading or unloading. The rollout roadmap starts with single-point pilots in biopharma, new energy and retail before expansion into production-line and campus-level workflows.
Source - 2026-06-30 - Funding - DexForce closes RMB1 billion Series B
New investors included Qianhai Mother Fund, Lens Technology, ICBC Capital, Hengjian Asset and Zhurui Capital, alongside returning shareholders. Proceeds were allocated to world-model algorithms, physics-engine upgrades, data infrastructure, robot development and real-world applications.
Source - 2026-06-02 - Capability - DexForce world model leads WorldArena Track 2 evaluation
WorldArena Track 2 evaluates world models as data-synthesis engines, policy evaluators and action-planning tools. DexForce adapted DexWorldModel from semantic-feature prediction to direct video generation for the Data Engine task. EmbodiChain supplies assets, scene layouts, trajectories and recovery data for the company’s world-model training pipeline.
Source - 2026-05-01 - Deployment - DexForce-powered service stations open across Chinese venues
Sites included Xuzhou Fantawild, Shenzhen Longgang Galaxy COCO Park, Shenzhen K11, Guilin Elephant Trunk Hill and Yangshuo West Street. The stations sold products including coffee, beer, ice cream, popcorn and soft drinks, using DexForce software to handle different containers and service equipment.
Source - 2026-02-28 - Partnership - DexForce partners with WeCook on robotic meal collection
The proposed workflow links personalized nutrition orders from the WeCook app to robot meal-selection and collection instructions. Robots would identify boxed meals in temperature-controlled cabinets and hand them to customers; the Hong Kong Productivity Council supported the collaboration.
Source - 2026-01-20 - Research Publication - DexForce open-sources EmbodiChain robot-training framework
Real2Sim extracts physical priors from limited real examples, while Gen2Sim builds three-dimensional scenes and assets from language instructions. The pipeline combines visual augmentation, randomized physics and automatic recovery trajectories after failed grasps; code and technical documentation were released publicly.
Source - 2025-07-10 - Funding - DexForce announces A1 and A2 financing rounds
Chengdu Science and Technology Venture Capital and Hongtai Fund led the financing, with new institutional participants and further investment from Lenovo Capital. Funding would support the Sim2Real model platform, X-Wiz software upgrades and expansion of the company’s robot product range and applications.
Source - 2025-01-20 - Product Launch - DexForce launches the DexForce W1 humanoid robot
DexForce W1 combines 34 powered units with seven-degree-of-freedom arms and six-degree-of-freedom dexterous hands. Its vision-based spatial sensors use multiple imaging modules and learned processing to identify objects under varied lighting. The DexVerse Sim2Real engine models lighting, friction and object dynamics to generate simulation data for robot training.
Source - 2024-05-14 - Funding - DexForce completes strategic financing led by Lenovo Capital
The company’s Sim2Real approach used generated three-dimensional data to train perception and manipulation systems for industrial robots. Its product portfolio combined the DexVerse simulation engine with vision-language-action perception tools for grasping and manipulation.
Source
Relations
Relationships
- Hung Fook Tong
Planned Food Service Concept - DexForce planned W1 Pro service robots for Hong Xiaofantang, a food-service concept created by WeCook and Hung Fook Tong.
- Bosuda
New Energy Factory Robotics - DexForce and Bosuda are developing an embodied-robot validation loop for new-energy manufacturing.
- Today International
Embodied Logistics Solution Partner - DexForce and Today International integrate embodied robots with warehouse systems and fleet scheduling for picking and transport.
- WeCook
Food Service Robotics Partner - DexForce and WeCook collaborate on W1 Pro robotic meal collection for WeCook and the Hong Xiaofantang food-service concept.
- Hong Kong Productivity Council
Food Service Project Supporter - The Hong Kong Productivity Council supports the DexForce and WeCook robotic meal-service collaboration.
Context
Comparable profiles
- Humanoid
comparable profile - Humanoid Industrial Builder · Wheeled Humanoid Builder
- Reflex Robotics
comparable profile - Wheeled Humanoid Builder
- Galbot
comparable profile - Wheeled Humanoid Builder
- WorkFar
comparable profile - Wheeled Humanoid Builder
- Figure AI
comparable profile - Humanoid Industrial Builder
- AI2 Robotics
comparable profile - Wheeled Humanoid Builder
- BlueBug Embodied (蓝虫具身)
comparable profile - Wheeled Humanoid Builder
- Ligong Industry (里工实业)
comparable profile - Wheeled Humanoid Builder
Shared workflows
- Holiday Robotics
workflow adjacency - Industrial Task Execution
- Agility Robotics
workflow adjacency - Industrial Task Execution
- Unitree
workflow adjacency - Industrial Task Execution
- Collaborative Robotics
workflow adjacency - Bimanual Mobile Manipulation
- Standard Bots
workflow adjacency - Industrial Task Execution
- Physical Intelligence
workflow adjacency - Embodied World Model Training
- Ropedia
workflow adjacency - Embodied World Model Training
- SI Robotics
workflow adjacency - Bimanual Mobile Manipulation
- OS3
workflow adjacency - Bimanual Mobile Manipulation
- ISYBOT
workflow adjacency - Industrial Task Execution
- Maven Robotics
workflow adjacency - Bimanual Mobile Manipulation
- Reactor
workflow adjacency - Embodied World Model Training
Context tags
- Embodied AI world models - Stack Layer - DexForce describes DexVerse, DexWorldModel and physical-AI world-model work as part of its core stack.
- Robot OEM - Stack Layer - Company builds DexForce W1 Pro humanoid robot platform.
- Wheeled humanoid builder - Peer Group - Official W1 Pro materials describe a wheeled humanoid robot with a humanoid upper body and mobile chassis.
- Humanoid industrial builder - Peer Group - Official W1 Pro scenarios include industrial manufacturing, data collection, commercial services and service automation.
- Mobile manipulation robotics - Stack Layer - W1 Pro is a wheeled humanoid platform with perception and manipulation, but use mobile_manipulation_robotics as supporting rather than sole identity.
- Bimanual mobile manipulation - Workflow - W1 Pro has dual 7-DOF arms, optional dexterous hands and a mobile chassis for manipulation tasks.
- Dexterous manipulation - Capability - W1 Pro supports dexterous hands, force control and dual-arm manipulation workflows.
- Industrial task execution - Workflow - DexForce product positioning targets industrial humanoid and automation scenarios.
- Vision-language-action control - Capability - DexForce positions its system around embodied intelligence, vision-first perception and Sim2Real/VLA-style robot learning.
- Factories - Customer Environment - DexForce names intelligent manufacturing and industrial production environments for W1 Pro and its physical-AI stack.
- Embodied world model training - Workflow - DexVerse and EmbodiChain materials describe synthetic data, simulation and model-training workflows for embodied tasks.
- Pilot deployments - Commercial Stage - Public materials report commercial and service deployments, but precise customer-scale robot counts should remain conservative.