Alibaba's Tongyi platform launched the universal embodied intelligence unified motion framework Qwen-VLA.

Tongyi Labs has launched Qwen-VLA, which uses Qwen3.5-4B as its base and combines it with a DiT motion decoder. Through a unified motion trajectory prediction framework, ontology-aware cueing conditionalization, and text-to-action (T2A) pre-training technology, it integrates manipulation, navigation, and trajectory prediction into a single, general embodied model. The model outperforms dedicated models in multiple benchmark tests and supports cross-body control across 11 robot platforms.