iFLYTEK-Embodied-Omni — unified vision, language, action model
iFLYTEK released a multimodal foundation model that unifies vision, language, and robotic action in a single framework instead of cascaded pipelines.
The model jointly processes videos, images, and text to generate control signals—avoiding the interface bottlenecks and error compounding of separate vision-reasoning and action-generation stages.
Designed for embodied agents that must interpret instructions, forecast scene evolution, and execute precise long-horizon tasks.