#China #Shanghai #AWE #TARS #Sail

TARS Shines at WAIC 2026

In a remarkable display of technological prowess, TARS celebrated its debut at the World Artificial Intelligence Conference (WAIC) 2026 held from July 17 to 20 at the Shanghai World Expo Exhibition & Convention Center. The highlight of TARS’ presentation was its AWE model (AI World Engine) for embodied artificial intelligence, which earned the prestigious SAIL Award (Superior AI Leader) recognizing its innovative technological advancements and substantial potential in various industries.

Dr. Chen Yilun, the founder and CEO of TARS, presented the latest iteration of the AWE model during the main forum. The AWE 3.5 is designed explicitly for embodied AI and is built upon over one million hours of real-world, human-centered data validated under industrial conditions. This revolutionary model integrates action, perception, geometry, and tactile sensing into a unified framework. Compared to its predecessor, Pi 0.5, AWE 3.5 enhances task execution efficiency by nearly two times, excels in complex task performance, and maintains a closed-loop interaction even for tasks extending several minutes.

Furthermore, AWE 3.5 is the first model to fully implement a 'Pre-training + Fine-tuning' paradigm specifically designed for embodied AI, enabling reproducible, scalable, and continuous development.

At the conference, TARS showcased the capabilities of AWE-powered robots in practical applications such as mobile phone packaging, backpack organization, and precise sorting of screws. These demonstrations illustrated how the model translates its understanding of the physical world into actionable tasks while adapting to unfamiliar objects and scenarios.

In a significant expansion of its capabilities, TARS has plans to boost its pre-training dataset from one million hours to ten million hours by the end of 2026. This increase aims to enhance generalization abilities, long-term reasoning, and the execution of real-world tasks, reinforcing the framework's robustness and versatility.

Additionally, TARS constructed a full-sized circular production line for automotive wiring harnesses, employing multiple A1 robots for gripping, routing, connecting, and assembling flexible wiring harnesses. This setup received special recognition as part of the WAIC exhibition area for embodied robotics, highlighted in the Smart Manufacturing Hub. The demonstration served to showcase the reliability, stability, and added value of this technology in advanced manufacturing environments.

A captivating feature of the event included a live demonstration of DexHand on an A1 robot in collaboration with magician Deng Nanzi, illustrating real-time perception, precision control, and trustworthy human-robot collaboration through activities like card spreading, handwritten notes, and even solving a Rubik's Cube.

Together, AWE 3.5, the industrial production line, and DexHand embodied TARS' tri-fold vision of using trustworthy physical AI to understand complex environments, create value in real production lines, and perform high-precision tasks. TARS is committed to advancing embodied AI from the demonstration phase to practical applications, further accelerating its large-scale deployment in the industry.

In conclusion, TARS’ impressive performance at WAIC 2026 marks a significant milestone in artificial intelligence's integration into real-world applications. With its groundbreaking technologies, TARS is not only shaping the future of AI but also setting new standards for reliability and efficiency in manufacturing and beyond.

Contact:
TARS PR Team
Email: [email protected]