Embodied AI startup PHYMI has raised nearly $100 million in its seed round. The round is led by IDG Capital, joined by Yunqi Partners, Didi, Fosun RZ Capital, Gaorong Capital and Hesai Technology. Founded in March 2026, PHYMI is headed by Liu Nianqiu, ex-technical partner and director at DeepRoute.ai. He took part in China’s high-level autonomous driving industry’s key journey from L4 R&D to mass production.
The 2026 China International Fair for Trade in Services (CIFTIS) opened in Beijing on Tuesday. Maniformer has featured at a special booth hosted by the Beijing Disabled Persons’Federation, themed “Digital Intelligence for Disability Assistance, Skills for a Better Life”. The display demonstrates new opportunities for people with disabilities to take part in robot training and gain income in daily life. At the exhibition booth, visitors with disabilities completed a series of tasks using two embodied intelligence data collection products, MEgo View and MEgo Gripper. Their movements are converted into datasets to train embodied AI models, unlocking a new occupation: robotics trainer.
Lightwheel and Peritas AI announced they have formed a strategic partnership to bring robots into perioperative healthcare workflows by combining infrastructure and orchestration technologies to enable deployment in clinical environments. According to the companies, the collaboration will integrate Peritas AI’s orchestration platform with Lightwheel’s physical AI infrastructure, allowing robotic systems to be trained, validated and operated within real-world hospital workflows. The companies said the multi-phase initiative, valued at about $56 million, could support deployment of up to 200 humanoid robots across 2026 and 2027. The companies noted healthcare operations, particularly in perioperative settings, remain highly fragmented, spanning sterile processing, operating rooms, logistics and equipment coordination across disconnected systems. The partnership is designed to address those gaps by enabling continuous, real-time...
Chinese humanoid robot maker UBTech said it has signed contracts with customers in major markets including Europe, Japan and South Korea, winning overseas orders worth more than 50 million yuan ($6.9 million). The orders include its commercial service humanoid robot Walker C1 and ultra-realistic humanoid robot U-World U1, according to UBTech.
Daimeng (Shenzhen) Robot Technology Co., Ltd. has undergone industrial and commercial changes recently. Shanghai Yunyang Enterprise Management Consulting Co., Ltd., an affiliate of Ant Group, has joined as a new shareholder. Meanwhile, the company’s registered capital has increased from roughly 1.917 million yuan to around 2.092 million yuan. Founded in December 2021 and headed by legal representative Duan Jianghua, Daimeng engages in R&D, manufacturing and sales of intelligent and industrial robots. The firm is now jointly held by Duan Jianghua, Aoli (Tianjin) Technology Partnership (Limited Partnership), and the newly added shareholder.
Chinese robotics company Unitree said it has fully open-sourced an upgraded version of its general-purpose humanoid foundation model, UnifoLM-WLA-1.0, setting new state-of-the-art results on multiple global open-source embodied model benchmarks. The model can perform both tabletop manipulation and whole-body mobile manipulation with a single model, enabling cross-task and cross-end-effector generalization, Unitree said. It achieves a “one-model-driven, whole-body coordination” effect, the company added. Unitree said UnifoLM-WLA-1.0 was trained on large-scale general-purpose multimodal perception and understanding datasets and incorporates interaction-centric world modeling, significantly strengthening spatial perception and scene understanding. It outperformed several open-source competitors in China and overseas on multiple embodied reasoning benchmarks, with some capabilities comparable to mainstream closed-source models, according to the company. In real-robot tests, a single model completed...
Unitree Robotics said it has fully open-sourced UnifoLM-WLA-1.0, an embodied foundation model that uses a single model to coordinate desktop and whole-body mobile manipulation, supporting cross-task and cross-end-effector generalization.
Ant Lingbo Technology said it has released three models: LingBot-World 2.0 Small (1.3B), LingBot-World 2.0 Bidirectional and LingBot-World 2.0 Causal Pretrain. The open-sourced LingBot-World 2.0 Small has 1.3 billion parameters, is designed for consumer-grade single-GPU use and can generate in real time on a single GPU.
DeepSeek officially released V4.1 Flash on Sept. 10, with China’s National Supercomputing Internet launching API services for the model the same day. The platform has connected more than 30 computing centers and nearly 70 specifications of computing resources. Its AI community hosts over 1,800 open-source large models and provides access to APIs from domestic models including GLM, Qwen and Kimi.
Chinese AI developer ModelBest said its MiniCPM5-2B model has ranked No. 1 on Hugging Face’s Trending list and first among global open-source models under 4 billion parameters on Artificial Analysis, with a composite score of 23. With 2 billion parameters, the model can handle agentic tasks including tool calling, deep search and code generation, ModelBest said. It marks an early prototype of a general-purpose on-device agent.
AGIBOT智元