Qwen Launches First Omni-Modal Model Qwen3.8-Omni-Flash with Agentic Capabilities

Alibaba's Qwen team released Qwen3.8-Omni-Flash on September 18, 2026, marking its first omni-modal model built around agentic capabilities. The model natively supports audio-video understanding, reasoning, and tool use, enabling tasks such as auto-editing vlogs, translating short videos, and generating movie recaps.

According to the official announcement, Qwen3.8-Omni-Flash approaches Gemini 3.8 Flash in audio-video capabilities and achieves an average improvement of 19.5 points in agent performance on WildClawBench-MM and UniClawBench. It features a 1M-token context, allowing active exploration of long videos and more accurate location of key moments, while using 51.8% fewer tokens than static understanding on OmniVideoBench.

The model also reduces video input costs by about 89% compared with Qwen3.5-Omni-Plus, making long-form audio-video understanding and agentic workflows more affordable. To support developers, Qwen has open-sourced Qwen-MM-Plugins and Qwen-Live Harness.