Galbot said its G1 robot can work in a variety of environments, including pharmacies. | Source: Galbot
Galaxy General Robot Co., also known as Galbot, this week completed a funding round of RMB 1.1 billion, or around $153 million. The company said this brings its total funding to about $335 million raised in the past two years.
In March, Galbot released its flagship robot, the G1. The semi-humanoid mobile manipulator features wheels and two arms. It is designed to automate inventory, replenishment, delivery, and packaging. The company said its artificial intelligence models enable G1 to handle 5,000 different types of goods.
Galbot said the robot only takes a day to deploy into a new store. Currently, nearly 10 stores in Beijing have deployed the robot. The company expects to deploy it into 100 stores across the country within the year.
Founded in May 2023, Galbot uses simulated synthetic action data sets for pre-training and real data for post-training. The Beijing-based company said this approach leads to the rapid iteration of embodied large model technology worldwide.
VLA models power G1 robot
Galbot said it has developed three proprietary vision-language-action models (VLAs). The first, released in January 2025, is GraspVLA, which the company described as an “end-to-end” embodied model that is pre-trained on simulated synthetic data.
Also, Galbot said it is working with NVIDIA Isaac Sim, and the model can achieve zero-shot generalization with only pre-training.
Next, it created GroceryVLA, a model for retail commercialization. Galbot said this model solves two key challenges in complex retail environments:
- Intelligent grasping of many objects: In a scenario involving complex and densely stacked goods, the model doesn’t need to adjust parameters for each product separately, Galbot said. Whether it be soft bags, hard boxes, bottles, glass containers, plastic packaging, or more, the model enables robots to accurately grasp the item, the company claimed.
- High efficiency in scene adaptation: Galbot said the model has strong autonomous decision-making capabilities, such as dynamically selecting the optimal operation target. It also has anti-interference capabilities, meaning it can deal with accidental displacement or dumping of goods. More importantly, it doesn’t require any scene pre-collecting to deploy.
Finally, Galbot offers TrackVLA, a product-level navigation model. The company said it can perform a range of tasks, including pure visual environment perception, comprehension of natural-language commands, zero-sample generalization in complex scenes, following highly dynamic targets, and more.
Galbot invests in collaboration
CATL and Puquan Capital led Galbot’s funding round. Other investors included China Development Bank, China Development Bank Science and Technology Innovation Fund, Beijing Robot Industry Fund, Qiming Venture Partners, and more.
CATL was the core strategic investor. The companies plan to collaborate on developing applications for Galbot’s general embodied intelligence large model and promoting it.
Galbot has also worked to strengthen its ties to the automotive industry. Last month, it announced a joint venture with Boyuan Capital, a subsidiary of the global industrial giant Bosch Group. The startup also signed a tripartite strategic memorandum with Bosch China and Boyuan Capital.
The partners said they will focus on industrial manufacturing and jointly promoting commercial applications for embodied intelligent robots.
Save now with early bird discount
Tell Us What You Think!