Overview
This is the real-time version of Qwen3-Omni-Flash multimodal large model, based on the Thinker-Talker Hybrid Expert (MoE) architecture. It supports efficient understanding and speech generation of text, images, audio, and video, enabling text interaction in 119 languages and voice interaction in 20 languages. It supports 49 voice timbres and generates human-like speech for accurate cross-language communication. The model features powerful command following and system prompt customization capabilities, flexibly adapting to dialogue styles and role settings. It is widely used in text creation, voice assistants, multimedia analysis, and other scenarios, providing a natural and smooth multimodal interactive experience. This version is a snapshot from December 1, 2025.
Input
Output
Features
Prefix Completion
Enable Partial Mode when calling the Qwen API to make the model continue strictly from your provided prefix text.View docsFunction Calling
Use function calling to connect large language models with external tools and systems.View docsCache
Context Cache stores shared prefixes for long-context requests to reduce repeated computation, improve latency, and lower cost.View docsStructured Outputs
Structured Outputs help ensure the model returns a JSON string in the expected format.View docsRate Limits & Context
- Max Input49K
- Max Output16K
- Max Input (Thinking)16K
- Max Output (Thinking)16K
- Context65K
- Max Reasoning32K
- TPMTokens Per Minute100K
- RPMRequests Per Minute60