Alibaba's Omni-Agent Just Crushed Video AI

🔥 What happened Alibaba dropped Qwen3.8-Omni-Flash, its first omni-modal model built for agentic work — text, image, audio, and video in, text out. It's API-only; no open weights at launch. 💡 Why it matters The agent doesn't watch whole videos anymore — it hunts for evidence. OmniVideoBench jumps from 63.4 to 67.8 while using 45.7% fewer tokens. Audio input costs over 98% less per hour than the previous model. ⚡ Our take The token savings are the real story, not the benchmark bumps. Anyone still feeding entire videos into a model is burning cash — and Qwen knows it.