Ant Financial's LingBot Technology launches LingBot-VA 2.0, the industry's first embody-based native world motion model.

Antminer Technology has launched LingBot-VA 2.0, the industry's first embodied native world action model. The model is pre-trained from scratch based on an autoregressive architecture and adopts four core designs: semantic vision-action segmenter, causal pre-training paradigm, MoE architecture, and asynchronous inference mechanism. It achieves 150Hz real-time inference on a single card, enabling robots to have general control capabilities that can infer and act at the same time.