The latest AI system from Alibaba integrates native audio-video reasoning and tool use to automate tasks such as editing vlogs and translating short videos. Qwen3.8-Omni-Flash matches Gemini 3.8 Flash in audio-video capabilities and posts a 19.5 point average increase in agent performance across WildClawBench-MM and UniClawBench. The model supports a 1M token context window for exploring long video files.

Video input costs for the architecture are approximately 89% lower than Qwen3.5-Omni-Plus. The model uses 51.8% fewer tokens than static understanding to locate key moments in video on OmniVideoBench. Alibaba is also open-sourcing Qwen-MM-Plugins and Qwen-Live Harness to assist developers in building applications around the omni-modal design.