Alibaba has introduced Qwen3.8-Omni-Flash, a multimodal model that can process text, images, audio and video with a 1-million-token context window. One of its key updates is a major reduction in audio and video processing costs. According to Alibaba, compared with Qwen3.5-Omni-Plus, the cost of processing one hour of audio has fallen by more than 98%, while combined audio-video processing costs have dropped by more than 93%. Instead of processing long audio or video files in full, Qwen3.8-Omni-Flash identifies the parts relevant to a user’s request and analyzes those fragments. This also helps reduce token usage.
According to Alibaba, the model improved by more than 25% on average across 29 benchmarks compared with the previous version. It supports audio in 113 languages and dialects. The model is available through an API and can be used for analyzing meetings, summarizing videos, generating subtitles, and building AI agents that work with audio and video. Alongside the new model, Alibaba also introduced Qwen-MM-Plugins and Qwen-Live Harness, tools designed to help developers build AI agents that can work with audio and video.
The use cases presented include video editing, music video creation, film analysis, short-series translation, and turning educational videos into structured notes. Qwen3.8-Omni-Flash is currently available via API. According to Alibaba Cloud’s documentation, the model also supports Chat Completions and Responses interfaces compatible with OpenAI APIs. For Alibaba, the update is not only about improving benchmark performance, but also about reducing the cost of deploying AI agents that work with audio and video at scale.
