Go1 Pro is an enhanced version of Go1 designed for consumer companionship, exploration and education scenarios. It weighs approximately 12 kg, equipped with five perception cameras, an adaptive payload of about 4 kg, a maximum speed of 3.5 meters per second, and adds God View plus upgraded perception configurations. ⚠️ This information is compiled from publicly available online sources, for reference only.
Unitree H2 is a full-size humanoid robot designed for industrial, logistics, research, and education applications. It measures 1,820 × 456 × 218 mm and weighs approximately 70 kg including the battery. The platform features 31 degrees of freedom, joints powered by low-inertia, high-speed inner-rotor permanent magnet synchronous motors, a stereo camera, a microphone array, a high-power speaker, Wi-Fi 6, and Bluetooth 5.2. It uses a quick-release smart battery and comes with a charger and a manual remote controller. The officially rated arm payload is approximately 7 kg, with a maximum of approximately 15 kg. ⚠️ This information is compiled from publicly available online sources, for reference only.
Zhejiang Humanoid Robot Innovation Center Co., Ltd. brought its NAVIAI humanoid robots to the 16th Smart City and Intelligent Economy Expo, which opened in Ningbo on Friday, showcasing real-world applications in apparel manufacturing and smart retail. The company said its robots are moving into real scenarios to perform real tasks, backed by a full-stack, self-developed product system. It is targeting two ends of the market: production sites that demand efficiency, precision, production-line compatibility and stable operation, and consumer settings that require natural interaction, environmental adaptability and flexible service. In apparel manufacturing, Zhejiang Humanoid has focused on flexible production, where soft fabrics are prone to wrinkling, collapsing and unpredictable deformation. Over two years, it said it moved from template-machine operations to...
At the JDDiscovery 2026 main forum, JD.com projected its five-year procurement targets on screen: 3 million robots, 1 million autonomous vehicles and 100,000 drones. The audience included more than 100 hardware manufacturers and dozens of investment firms, many expecting humanoid robots to feature on the list. None appeared. On exhibition zones flanking the stage, JD’s Wolf-series machines handled warehouse tasks: Dilang shuttled goods between racks, Canglang picked orders in narrow aisles, and Dulang completed campus deliveries. Not a single bipedal robot was on display. JD unveiled six new Wolf-series robot models at the event. The omission of humanoids was no oversight. With China’s largest fulfillment network, JD’s procurement blueprint signals a clear verdict: humanoid robots are not yet commercially viable...
The Chinese Academy of Science and Technology for Development unveiled its 2026 Report on China’s Regional Science and Technology Innovation Evaluation at the Pujiang Innovation Forum. The report puts China’s overall score for scientific and technological innovation at 81.13, up 0.93 points year-on-year, reflecting steady improvements in the innovation ecosystem. Shanghai, Beijing, Jiangsu, Guangdong, Zhejiang and Tianjin lead the nation in comprehensive innovation capacity. With a score of 91.40, Shanghai keeps its No.1 position after claiming the top spot last year.
Robot training data startup XDOF is in late-stage talks for a Series B round at a valuation of about $1.2 billion, led by 8VC, just three months after closing a $70 million Series A. Annualized revenue has approached $50 million, serving about 20 clients, including several frontier AI labs. Founded in 2024 by UC Berkeley researchers Philipp Wu (CEO) and Fred Shentu (CTO), XDOF collects real-world teleoperation data to train general-purpose robots. Unlike large language models, robots have no internet-scale off-the-shelf data to draw on. XDOF’s business is to fill that gap.
GenRobot.AI is spearheading a new paradigm of whole-body data with the release of its Whole-Body Manipulation Data technology. The innovation enables robots to master full-body coordination, accelerating the advent of Physical AI.
What happens if robot model training data is expanded from 300 hours to 30,000 hours? Can robotic systems, similar to large language models, continuously gain new capabilities as datasets grow? AGIBOT’s newly released native World Action Model (WAM), GE-Act 2.0, systematically answers that question for the first time. Unlike mainstream approaches that build upon existing video generation models, every parameter of GE-Act 2.0 — including core modules for visual representation, future prediction and action forecasting — is initialized from scratch and trained end-to-end on embodied manipulation data. After pre-training, the model undergoes zero-shot real robot tests on unfamiliar scenes and unseen objects without fine-tuning or demonstration for benchmark tasks. The evaluation covers 100 atomic tasks, 20 skill categories and two...
The awakening of robotic intelligence may already be underway. AgiBot has released GE-Act 2.0, its upgraded world action model, featuring a 100-fold expansion of training data that enables robots to independently master new capabilities. The new iteration drastically scales the volume of training corpus, empowering robotic agents to discover and pick up new skills without manually coded instructions, marking a notable leap for embodied intelligence.
Maven Robotics, a California-based startup developing general-purpose robots for industrial work, said on Thursday it raised $100 million in Series A funding to expand its logistics and manufacturing platform. The round included RoboStrategy, LocalGlobe, Vine Ventures and XTX Ventures, among others. Maven will use the proceeds to scale deployments and advance its system as it moves beyond palletizing and tote handling into broader industrial tasks. The company cited U.S. labor shortages expected to worsen as demand grows. It says its platform combines hardware, software and AI across robot, fleet and factory levels, and connects deployment data through a Maven network designed to share learning across systems. It targets industrial applications requiring reliable performance over multiple shifts. CEO Hamza Derbas, who...
AGIBOT