Qwen3.8-Omni-Flash launched with 1M context supporting text image audio video. It boosts Agentic workflows for editing and production. Scores rose over 26% versus Qwen3.5-Omni-Plus with significant gains in audio visual agent and coding tasks. Audio video API prices dropped over 93%. The model matches Gemini3.8Flash in audio video while exceeding it in overall audio. Token consumption reduced by 45.7% via Agentic Understanding mode. The team used the model for its own R&D completing four iterations in 12 hours. A Realtime version supports sound localization and improved Sichuan dialect recognition accuracy by 40.7%.