On July 17, 2026, Ali Tongyi Lab proudly announced the release of its latest innovation—the Wan-Streamer v0.2 model. This cutting-edge model is built upon a native streaming architecture and incorporates a distributed inference topology. It seamlessly synchronizes understanding and generation processes within a single model interaction loop, enabling end-to-end full-modality interaction. Specifically tailored for real-time duplex interaction, this model boasts an impressive end-to-end interaction latency of around 0.55 seconds (inclusive of network transmission time). Additionally, its output resolution has been upgraded to 640×368@25FPS, providing robust support for the real-time, synchronous understanding and generation of text, audio, and video content.
