Ant Group’s Robbyant Open-Sources LingBot-World for Real-Time AI Interaction
Event summary
- Robbyant, an Ant Group subsidiary, open-sourced LingBot-World, a world model enabling millisecond-level real-time interaction.
- LingBot-World achieves up to 10 minutes of continuous, stable video generation with 16 FPS throughput and sub-second latency.
- The model supports zero-shot generalization, requiring only a single real-world image or game screenshot for interactive video stream generation.
- Robbyant combines web videos and game-engine synthetic data for training, leveraging Unreal Engine pipelines for clean, UI-free frames.
The big picture
Ant Group’s open-sourcing of LingBot-World underscores the growing emphasis on embodied AI, where digital models interact seamlessly with physical environments. This move aligns with broader industry trends toward real-time, high-fidelity simulation for training autonomous systems. The release also highlights Ant Group’s push to extend its AI capabilities beyond digital finance into robotics and physical-world applications.
What we're watching
- Adoption Pace
- How quickly developers and enterprises integrate LingBot-World into autonomous driving and game development workflows.
- Competitive Response
- Whether rivals like NVIDIA or DeepMind accelerate their own world model offerings in response.
- Strategic Alignment
- The extent to which LingBot-World advances Ant Group’s AGI strategy into physical-world applications.
Related topics