The Qwen team launched Qwen3.8-Omni-Flash, a native omnimodal model that accepts text, image, audio, and video with a 1M-token context window, available on the Qianwen AI Platform. The team reports an average gain of more than 25% over Qwen3.5-Omni-Plus across 29 evaluations and audio input prices lower by more than 98% per hour.
On OmniVideoBench, an agentic mode that chooses which segments to watch raised accuracy from 63.4 to 67.8 while cutting tokens per query by about 45.7%. The team also open-sourced Qwen-Live Harness for real-time interaction. The post is marked as a draft and does not establish open model weights or independent results.