On June 17, 2026, the field of embodied intelligence ushered in a major breakthrough.
Quic Robot’s self-developed technology has successfully entered L’Oréal’s global strategic production base – Suzhou production plant. This not only proves that domestically produced lightweight VLA embodied intelligence products are fully capable of meeting the stringent standards of international high-end manufacturing, but also marks the first Chinese company to achieve a large-scale commercial breakthrough in lightweight VLA technology, becoming an important milestone for China’s embodied intelligence industry to move from technology research and development to large-scale and global commercialization.
Relying on the lightweight VLA model EVO that topped the Meta-World list –
1. Qingcang Robot only takes 30 days to complete the entire business process of scene data collection, model incremental adaptation, and on-site deployment and joint debugging.
This practice completely rewrites the long deployment cycle of 3 to 6 months for traditional industrial robots. It relies on the four core advantages of lightweight models, low computing power thresholds, high cost performance, and rapid adaptation to create a new industry paradigm of industrial embodied intelligence.
Facing the effect of the production line deployment, the relevant person in charge of L’Oreal Suzhou factory commented: “Qingcang robots and our innovative practice in Suzhou factory have once again confirmed the great value and development trend of artificial intelligence deeply empowering industrial production lines. L’Oreal has always actively embraced technological changes, and we chose Qingcang as the embodied intelligence partner of the core production line. Its efficient delivery capabilities, solid scene research capabilities and deep technical accumulation are impressive.
I am very impressed.
We look forward to using this cooperation as a starting point to continue to explore application scenarios and expand deeper cooperation at more levels.”
As the core base of L’Oréal’s global supply chain, the Suzhou factory is responsible for the high-end production of full-line cosmetics. It has a rich variety of materials on site, large differences in product forms, strict operating accuracy requirements, and dynamic and changeable operating environments, which puts forward extremely high requirements for embodied intelligence’s scene understanding, dynamic adaptation, and stable operation capabilities.
Under the traditional industrial automation model, robots are highly dependent on customized program development and special tooling fixtures. Not only does the deployment cycle take 3 to 6 months, but the overall cost of hardware investment and customization remains high, and it is difficult to adapt to multi-category and flexible production scenarios. It has long been a core constraint on the intelligent transformation of flexible manufacturing industries such as beauty and daily chemicals.
Taking a direct hit on the industry’s pain points, Qingcang Robot abandons the long-standing split technology architecture of “perception + planning + control” in the industry, and independently develops an end-to-end lightweight VLA embodied intelligent base model to achieve direct generation from multi-modal original input to the robot’s continuous action sequence. This is also the core support for the project to achieve extremely fast and low-cost implementation.
After in-depth architecture optimization, parameter simplification and training paradigm innovation, this model has achieved the ultimate lightweight of 0.8B parameters while retaining the full-dimensional multi-modal understanding ability. It can complete real-time inference operations by relying only on the conventional computing power of the robot end side, significantly reducing hardware investment costs.
At the same time, the model natively supports rapid adaptation of small samples. When faced with new operating scenarios, it only needs to collect a small amount of on-site real data to complete incremental tuning, and then it can quickly adapt to production line working conditions, eliminating the tedious links such as scene modeling, hardware modification, and repeated debugging in traditional solutions.
With this leading lightweight VLA technology system, Qingcang Robot has significantly compressed the overall closed-loop deployment cycle of industrial robots from the industry average of 3-6 months to within 30 days, achieving a leap-forward improvement in deployment efficiency.
After entering the core production line, Qingcang wheeled dual-arm humanoid robot fully utilized the technical advantages of the lightweight VLA model and delivered outstanding results.
As the “intelligent brain” of the robot, the VLA model deeply integrates the three major modes of vision, language, and movement. It can accurately analyze complex on-site environments, independently identify material shapes, placement status, and production line operating conditions. It can immediately capture problems such as material misalignment, stacking abnormalities, and production line flow lags, and independently complete research, early warning, and emergency response. Core indicators such as abnormality recognition accuracy, response speed, and treatment effectiveness have successfully passed L’Oréal’s high-standard acceptance.
Relying on the powerful generalization ability of the model, the robot can seamlessly connect multiple mixed tasks such as dynamic replenishment, precise picking, flexible tallying, and mixed sorting of multi-category finished products, perfectly matching the changing operational needs of the production line.
Compared with the limitations of traditional automation equipment that can only adapt to standardized materials and fixed processes, the Qingcang embodied robot truly realizes one model to adapt to multiple scenarios and one model to cover multiple categories, completely solving the automation problem of high-end beauty flexible production lines.
The successful deployment of this project is not only a strong evidence of Qingcang Robot’s comprehensive technical strength and project service capabilities, but also opens a new chapter in the in-depth strategic cooperation between the two parties.
The two parties will gradually promote the lightweight VLA intelligent robot solution to more L’Oréal production and warehousing scenarios to help industry manufacturing achieve full-link intelligent upgrades.
Behind the extremely rapid deployment results, Qingcang Robotics has long insisted on self-research of full-stack technology and the deployment of high-end talents.
The company’s core team brings together the world’s top technology companies and senior experts in the field of head-on autonomous driving. They have cutting-edge technology research and development vision and front-line practical experience. They are deeply involved in the two core directions of lightweight VLA base models and low-cost implementation of robots. They have built an integrated professional system of “cutting-edge algorithm research and development + real-life scene implementation” to focus on creating embodied intelligent products with low computing power and high adaptability, which accurately meet the core demands of the global manufacturing industry for cost reduction, efficiency improvement, and flexible production.
Since its establishment, Qingcang Robot has always adhered to commercial implementation and technological innovation as its core driving force, focusing on the two major tracks of logistics and warehousing and pan-industrial manufacturing. With wheeled two-arm humanoid robots as its main products, relying on its exclusive lightweight VLA technology barriers, it has continuously broken through the capabilities of traditional automation and expanded the application boundaries of industrial embodied intelligence.
At present, industrial intelligence and unmanned operations have become an inevitable trend in the transformation and upgrading of the global manufacturing industry. Embodied robots, as the core equipment of a new generation of industrial intelligence, are facing broad development opportunities.
Facing the future, Qingcang Robot will continue to adhere to technological innovation, relying on the leading lightweight VLA basic model, full-stack technical system, rich implementation experience and professional talent team, and continue to focus on core tracks such as logistics warehousing and pan-manufacturing.
While helping domestic companies build fully unmanned warehouses and smart factories, we are steadily accelerating globalization and working with partners to carry out joint innovation.
The company will continue to use its leading lightweight VLA hard-core technology as its foundation to lead the logistics, warehousing and pan-manufacturing industries into a new era of Physical AI, and continue to contribute to the intelligent upgrading of the industry.
