QianWen has officially rolled out Qwen3.8-Omni-Flash, its latest native full-modal model. This cutting-edge model accommodates a wide array of input types, including text, images, audio, and video, and is capable of processing extensive 1M-token contexts. It boasts autonomous task-planning abilities and can leverage various tools to execute creative endeavors.

Building on its predecessor's strengths, Qwen3.8-Omni-Flash has significantly enhanced its audio and video processing prowess. It can now effortlessly manage several hours' worth of audio and video content, perform tasks like summarizing meeting minutes, and is well-suited for audio and video creation scenarios, such as music video production and short drama translation. Additionally, it can transform lengthy videos into valuable information assets. Its real-time interaction version is particularly well-suited for scenarios such as oral practice companionship.

In tandem with the model's launch, the official team has also expanded the functionality of Qwen-MM-Plugins and open-sourced Qwen-Live Harness. In comparison to its predecessor, the new model has achieved an average score improvement of over 26%. Moreover, the API pricing for audio and video inputs has been substantially reduced. Presently, users can access this advanced model on the QianWen AI platform.