Qwen3.8 Omni Flash Realtime
This real-time Omni model accepts streaming audio and video frames and returns text and speech. It supports tools and remote MCP services.
Technical details
WebSocket audio can use one, two or four channels. Media-history limits do not specify the total session duration.Qwen3.8 Omni Flash Realtime
Practical use
Practical guidance. Store important conversation facts explicitly rather than relying on an unlimited audio history. Check interruption behaviour and tool completion before using a spoken confirmation to report that an action succeeded.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | API |
| Model identifiers | qwen3.8-omni-flash-realtime |
| Regions | Beijing and Singapore; WebSocket, WebRTC or AOQ.Qwen3.8 Omni Flash Realtime |
| Retained media | Up to 100 audio turns or 600 audio seconds; up to 50 video turns or 240 video seconds.Qwen3.8 Omni Flash Realtime |
Related models
- Qwen3.8 Omni Flash
Alibaba · Qwen Omni
Multimodal model that returns text from audio, video, image and text inputs, with 113 audio languages and dialects.
Launch date not confirmedAPI - Qwen3.8 LiveTranslate Flash Realtime
Alibaba · Qwen LiveTranslate
Live translation model that understands 60 languages and produces speech in 29, with optional image input.
Launch date not confirmedAPI