Ant Lingbo Releases Industry’s First Embodied-Native World Action Model: LingBot-VA 2.0

Bot Telegraph, July 10 — Ant Lingbo today released LingBot-VA 2.0, the industry’s first embodied-native world action model. This model represents a critical architectural choice in embodied AI development: rather than grafting capabilities from digital world models, the robot’s “brain” is natively designed from the ground up for dynamic modeling, causal prediction, and real-time execution in physical interaction with the environment.

According to the company, LingBot-VA 2.0 works in synergy with the previously released LingBot-World 2.0 world model. The Lingguang App has already integrated these capabilities — users can not only explore AI-generated scenes but also perform skill-based interactions similar to a 3D game, with AI generating real-time scene feedback based on user actions.

The core breakthrough of this technical approach lies in shifting the embodied AI “brain” from digital-world simulation to physical-world native interaction, which is expected to significantly improve robot adaptability and operational precision in real-world environments.