Give robots "wings of imagination"

If one day, your home robot no longer just mechanically follows commands, but can anticipate that a cup might be knocked over when helping you pour water and adjust its angle in advance; when you stand up, it can “guess” what you want to pick up and proactively hand it to you… What does this mean? The answer is: robots are beginning to have “imagination.”

How can robots acquire such “imagination”? Haidian embodied intelligence company Jihua Vision has provided its answer. Recently, this company, regarded by the capital market as the “OpenAI of the physical world,” announced it has completed nearly 1 billion yuan in Pre-Series B funding. Its core goal is to create a “brain” for robots that can “imagine” the future, promoting general-purpose robots into thousands of industries and millions of households.

Enabling robots to “predict” the future

When robots are no longer just moving mechanically according to preset programs, but can, like humans, anticipate environmental changes before acting and adjust their posture and actions accordingly, the core of this ability is “embodied schemas” and “world models.”

Jihua Vision founder and CEO Huang Guan used a vivid analogy: “Embodied schemas are like the ‘brain’ of the robot, and the ‘world model’ is the ‘imagination’ of that brain.”

Traditional robots rely heavily on real data for learning, which is inefficient and often leaves them “at a loss” in a constantly changing real-world environment. Jihua Vision’s breakthrough lies in not only developing industry-leading embodied schemas but also being the first in China to layout a “world model.” This world model functions like an efficient “data engine” and “thinking simulator,” allowing robots to learn and self-evolve in virtual worlds at 10 to 100 times the efficiency, significantly reducing the costs and time of real data collection. In short, Jihua Vision is enabling robots to go from merely “seeing” what’s in front of them to “predicting” the future, making smarter and safer decisions.

On the technical front, Jihua Vision’s GigaBrain series embodied schemas are at the global forefront, excelling in open-source schemas, foundational model reports, complex task performance on real machines, and large-scale authoritative real-machine evaluations.

Additionally, Jihua Vision has deep expertise in both model architecture and data engine development. In terms of architecture, the GigaBrain-0.5M* is the world’s first embodied schema based on a world model that achieves efficient reinforcement learning and self-evolution. They also released GigaWorld-Policy, which realizes comprehensive breakthroughs in real-time performance, training efficiency, and success rate for the “world-action model.” In data engine development, GigaWorld-0 is introduced, enabling the embodied world model to play a core role on high-level embodied schemas.

The development of embodied schemas also requires collaboration with native entities. It is learned that Jihua Vision released its new full-stack self-developed physical AGI native entity MakerH01 in November 2025, which has now entered large-scale mass production and delivery, targeting scenarios such as data collection, industrial applications, and services. In terms of generalization scenarios, Jihua Vision has established close cooperation with leading clients across automotive manufacturing, 3C electronics, warehousing logistics, guided tours, and home services.

Accelerating on the fertile ground of technological innovation

Jihua Vision’s rapid growth is inseparable from the fertile ground of scientific and technological innovation where it was born and developed. It is understood that Jihua Vision closely relies on the Intelligent Vision Laboratory of the Department of Automation at Tsinghua University. Its founder and CEO, Huang Guan, holds a PhD from Tsinghua’s Department of Automation, previously served as the head of visual perception technology at Horizon Robotics, and as a partner at JianZhi Robotics. He has experience working at Microsoft Asia Research Institute and Samsung China Research Institute, has led teams to win global AI competition championships multiple times, and has published numerous internationally renowned AI results, with rich experience in technological innovation, industry implementation, and continuous entrepreneurship in physical AI.

Moreover, the core team combines cutting-edge innovation in physical AI with full-stack mass production experience. It includes former vice presidents and architecture experts from major internet companies, as well as Huawei “Genius Youth” awardees; the native entity R&D team comes from leading domestic robotics companies, having led the development and delivery of thousands of humanoid robots, known as the “Dream Team” of physical AGI foundational models.

“From the scientific research achievements in Tsinghua labs to becoming an industry leader today, we deeply feel the strong support from Haidian District in technological innovation and industrial transformation,” Huang Guan said. Haidian District boasts a dense concentration of top universities and research institutes, providing continuous intellectual support for enterprises. Its complete industrial ecosystem and forward-looking policies create the best environment for original breakthroughs from 0 to 1 and industrialization from 1 to 100.

Riding the wave of the physical world with capital — “GPT-3 moment”

Jihua Vision’s nearly 1 billion yuan Pre-Series B financing features a “star-studded” lineup of investors. According to reports, investors include top chip and automotive industry capital such as SMIC Juyuan, Shanghai Semiconductor Industry Investment Fund, Linxin Capital, Xingyuan Capital, Wanlin International, as well as state-owned capital and financial institutions like China International Capital Corporation (CICC), SuChuang Venture Capital, Huaqiang Capital, and Yangtze Capital. Notably, CICC, Huaqiang Capital, Caixin Capital, and Zhang Keya Kun, as longstanding shareholders, continue to provide substantial support.

After the funding, Jihua Vision’s shareholder structure has formed a “combination of industrial capital + state-owned platforms + financial institutions.” This diversified capital structure not only offers strategic synergy and industrial resources but also provides long-term stable financial backing and regional ecological empowerment.

Huang Guan revealed that the company has now formed a full-stack layout of “embodied schemas — world models — native entities — generalization scenarios.” Its self-developed physical AGI native entity “MakerH01” has begun large-scale mass production and delivery, aiming for thousands of units this year, with pioneering applications across automotive manufacturing, 3C electronics, and home services.

He added that this round of financing will be used to continue leading the evolution of embodied schemas and world models, while advancing the adaptation of native entities and the deployment of generalization scenarios, striving for the “GPT-3 moment” in physical world AGI with maximum efficiency, speed, and performance.