Raised 500 million yuan in 3 months, the embodied intelligence company "Jijia Vision" completed a 200 million yuan Series A2 financing and launched the native model and native ontology of physical AGI.

Just one month after the announcement of the previous round of financing, the embodied intelligence company "Jijia Vision" has recently received a new round of investment.

It is reported that this round is a 200 million yuan Series A2 financing. It is led by Fortune Capital, co-led by the old shareholder Huakong Fund, and followed by well-known institutions such as Shoufa Development Venture Capital, Puyao Xinyie, Caixin Capital, Zhuhai Science and Technology Industry Group, Zhangke Yaokun, and Fuzhuo Venture Capital. The old shareholder Hedinggong Capital has oversubscribed.

Previously, Jijia Vision had completed three consecutive rounds of financing, namely Pre-A, Pre-A+, and A1. It completed a total of 500 million yuan in Series A financing in four rounds within three months.

As a company focusing on general intelligence in the physical world, Jijia Vision has not only launched a native model targeting physical AGI but also released the corresponding ontology on November 26, 2025, to layout the terminal business of physical AGI.

Specifically, the company's products include the world model platform GigaWorld (for driving and embodiment), the general embodied brain GigaBrain, and the general embodied ontology Maker, which are full-stack hardware and software products for physical AI. The product matrix systematically lays out the future development path of physical AI from both hardware and software ends.

In terms of models, Jijia Vision has proposed a native paradigm of "world model + action model + reinforcement learning", in which each link is driven by the world model.

Currently, the model architecture is converging towards general action models (such as VLA and world action models); the data sources are shifting to real - machine data and data generated by the world model as the core; the learning method has formed a general paradigm of "imitation learning + reinforcement learning + world model". It can be said that the world model is becoming the core driving force for physical AGI in terms of data sources, learning methods, and model architecture.

In view of the importance of the native model, the native ontology centered on "operation and upper limbs", which can better interact with the physical world and prioritize data, is increasingly becoming a key requirement. The scalable closed - loop iteration among "sensors - actuators - data acquisition devices - general models" is also becoming more and more valuable.

Jijia Vision believes that physical AI is entering a new critical era, and the next 2 - 3 years are the critical window period for the breakthrough of physical AGI. With the continuous breakthroughs in the world model and action model, the "ChatGPT moment" in the physical world is accelerating!

The in - depth combination of the native model and the native ontology is the key to achieving the above goals.

The World - Champion Team Deeply Engaged in Physical AGI

Looking back at past achievements, Jijia Vision has released and open - sourced the world's first world - model - driven embodied VLA large model GigaBrain - 0 and the industry - leading embodied world model GigaWorld - 0.

The data in the previous article "Jijia Vision Receives 100 - million - yuan Series A1 Strategic Financing Co - led by Huawei Hubble and Huakong Fund, Leading the Physical AI to the Final Technical Route with the 'World Model'" shows that the company's model performance has reached the advanced level in the industry.

In addition, the company has recently released its self - developed new - generation physical AGI native ontology Maker H01. It has also started large - scale mass production and delivery.

Jijia Vision closely relies on the Intelligent Vision Laboratory of the Department of Automation at Tsinghua University. The core team has witnessed the development of physical AI in the past decade and has continuously achieved world - class results in technological innovation and industrial implementation.

In the CV era, the team has repeatedly led to win world championships in the world's most influential visual AI competitions such as FRVT, COCO, and VOT, and has achieved large - scale industrial implementation.

In the era of autonomous driving, the series of works such as BEVDet proposed by the team are one of the most influential autonomous driving BEV paradigms in the world, ranking first in nuScenes globally for a long time, and have achieved large - scale mass production and implementation. The team has also led the R & D and commercial implementation of the industry's largest - scale data closed - loop products such as the Horizon AIDI platform and the R & D and commercial implementation of the Baidu Apollo simulation framework.

In the era of world models and embodied intelligence, Jijia Vision's DriveDreamer series is one of the world's earliest representative works on world models for the physical world and has achieved large - scale mass production and implementation. GigaBrain - 0 is a self - developed embodied VLA large model at the leading level in China driven by the world model. GigaWorld - 0 is the world's first systematically introduced embodied world model, whose value has been verified on a large scale on real machines and is widely used in real - world scenarios.

Dr. Huang Guan, the founder and CEO of Jijia Vision, is an innovative leading engineering doctor from the Department of Automation at Tsinghua University. He has served as the head of vision perception technology at Horizon Robotics, a partner and vice - president of algorithms at Jianzhi Robotics, and has worked in top research institutions such as Samsung China Research Institute and Microsoft Research Asia.

Dr. Huang Guan has witnessed the technological and industrial development of physical AI in the past decade and has led the team to continuously achieve breakthrough global leading results. He has leading - edge technological innovation, industrial implementation, and continuous entrepreneurship experience in the field of physical AI.

Physical AGI Native Model

The "physical AGI" proposed by Jijia Vision refers to general artificial intelligence capable of interacting with the physical world. Currently, the company has released the physical AGI native action model GigaBrain - 0 and the physical AGI native world model GigaWorld - 0.

GigaBrain - 0 is the brain of the embodied robot, providing end - to - end decision - making and control. By using the current most advanced VLA model architecture, GigaBrain - 0 supports the input of images, depth, text, and ontology states, outputs structured task planning and motion planning, and has completed pre - training on a large amount of data.

To address the problems of insufficient accuracy and reasoning ability in current robot operations, GigaBrain - 0 has also focused on improving 3D spatial perception ability and structured reasoning ability, enabling it to complete more accurate mobile navigation and longer - sequence complex operations, and having stronger generalization ability.

Moreover, GigaBrain - 0 can complete multiple long - range complex tasks, including dexterous manipulation tasks (such as making coffee, folding cartons, folding clothes, and organizing toilet paper), long - range tasks (such as tidying up the desktop and pouring drinks), and mobile manipulation tasks (such as carrying boxes and picking up clothes). The success rate of GigaBrain - 0 is significantly better than the existing SOTA methods.

GigaWorld - 0 is the nourishment for the embodied robot, providing high - quality, large - scale training data.

Jijia Vision's self - developed "world model platform GigaWorld" generates high - fidelity, controllable, and diverse embodied interaction data through geometrically consistent and physically accurate world model modeling, realizing "data amplification" and constructing the most complete embodied data system. This enables the trained VLA model to achieve a nearly 300% performance improvement in three generalization dimensions: new textures (material textures not seen during training), new perspectives (observation angles not seen during training), and new object positions (spatial layouts not seen during training).

On the PBench (Robot Set) benchmark, GigaWorld - 0 was comprehensively compared with the current advanced world models and is the world model with the smallest number of parameters and the highest overall performance among all the compared models. This result proves that GigaWorld - 0 combines excellent generation quality and extreme reasoning efficiency in embodied intelligence tasks and is currently the most cost - effective world model solution.

△ Performance of Jijia Vision's model. Image source: Jijia Technology GigaAI WeChat official account

Physical AGI Native Ontology: Maker H01

For physical AGI, the native action model and the native world model alone are not enough. Jijia Vision has also launched the native ontology - Maker H01.

△ Jijia Vision's Maker H01 robot. Image source: Jijia Technology GigaAI WeChat official account

Maker H01, paired with a series of toolchains of GigaBrain and GigaWorld, can quickly achieve implementation and task generalization in open scenarios such as households, commercial services, and light industries at a lower cost and with a higher success rate.

This high - degree - of - freedom embodied intelligent robot adopts a dual - arm + mobile chassis form and can perform tasks such as picking and placing items, inspection, reception, and assisting experiments in various scenarios such as households, industries, services, and scientific research.

Its chassis is omni - directional and fully driven, with high mobility and endurance. Each arm has a kilogram - level load capacity and is equipped with multi - fingered grippers. The overall is more of an "engineering - available" embodied platform, which can play dual roles of continuous operation and algorithm verification in real business environments.

The robot is available in an adult - height version of about 1.6 meters, suitable for different space and application requirements. It has more than 20 degrees of freedom throughout the body and, combined with various sensors, can complete delicate operations and compliant motion control.

At the same time, Maker H01 supports multiple remote - operation schemes. Combined with the self - developed embodied brain "GigaBrain", it can achieve whole - body coordination in complex actions such as grasping, carrying, and cooperation.

With the continuous evolution of the native model and the native ontology, Jijia Vision has reached signing and mass - production cooperation agreements with multiple leading customers in the industry, including multiple leading global automakers, multiple humanoid robot innovation centers and training fields in different regions, and multiple leading companies in the embodied field for industrial, service, and household scenarios.

In the future, Jijia Vision will continue to iterate on the physical AGI native model of "world model + action model + reinforcement learning", accelerate the R & D of the physical AGI native ontology suitable for the embodied brain, and promote the implementation of benchmark scenarios and the layout of general scenarios. Through the trinity of "native model - native ontology - general scenarios", it will accelerate towards the "ChatGPT moment" of physical AGI!

This article is from the WeChat official account "Intelligent Emergence". It is published by 36Kr with authorization.