Alibaba’s Qwen team released Qwen3.8-Omni-Flash on September 18, 2026, through the Qianwen AI Platform. The model accepts text, images, audio, and video input with a 1 million-token context window. Native output is text; media production requires external tools via function calling.
Alibaba pairs the model with Qwen-MM-Plugins for multimodal agent workflows and Qwen-Live Harness for continuous audiovisual interaction. International pricing starts at $0.15/M input tokens with aggressive cache-hit discounts for repeated media analysis.
The release positions Qwen3.8-Omni-Flash as a lower-cost sensory layer for agent applications processing meetings, recordings, screens, and video before taking action. Available via Alibaba Cloud in six regions.
Parameter count was not disclosed. Alibaba claims 25%+ average improvement over Qwen3.5-Omni-Plus across 29 benchmarks (vendor-reported).