ITHome reports AgiBot launches WITA-Omni embodied multimodal interaction model

According to IT之家, AgiBot announced the release of WITA-Omni on November 10. The company says it is the industry’s first embodied end-to-end multimodal interaction model, trained on robot body perception data. It adds touch and internal-state inputs alongside audio and video, and can output actions and facial expressions. Built on a streaming full-duplex architecture, it reportedly responds in about 400 milliseconds at best and 600 milliseconds on average, while supporting free switching across hundreds of languages and dialects.