By No Code MBA · Facts checked
- Provider
- Qwen
- Announcement date
- Status at last check
- Generally available
- Model family
- Qwen3.8
- Capabilities
- Text, Image, Audio, Video
What changed
Qwen3.8-Omni-Flash is an omnimodal model with a 1M-token context that accepts text, images, audio (up to 3 hours) and video (up to 2 hours) and outputs text, with a separate realtime variant that also speaks. It supports function calling, web search and reasoning effort settings. Alibaba reports speech recognition in 74 languages and 39 Chinese dialects, a 25%+ higher average than Qwen3.5-Omni-Plus across 29 evaluations, and hourly audio input costs more than 98% lower than that model.
Access and availability
Call qwen3.8-omni-flash through Alibaba Cloud Model Studio's Chat Completions or Responses APIs, or qwen3.8-omni-flash-realtime over WebSocket or WebRTC, with Beijing and Singapore endpoints. It is also on the Qianwen AI Platform.
Official access documentationWhat you can build
Test the model on a multi-step video editing or meeting summarization workflow to evaluate its omnimodal planning and tool use reach.
Limitations
The non-realtime model outputs text only; spoken output requires the realtime variant. Open weights were not announced, and benchmark claims are Alibaba's own.