- Published on
Qwen-Audio-3.0 Series Launch on Alibaba Qwen AI, Leading Globally in Recognition, Synthesis, and Real-Time Interaction
- Authors
- Name
- DeepHorus
Alibaba Cloud announced today that its self-developed Qwen-Audio-3.0 series of speech models have officially launched on the Qianwen AI platform and Alibaba Cloud Bailian platform. Developers can access them via API or Token Plan. The series includes three models for speech recognition, speech synthesis, and real-time interaction, and swept first place in all three tracks on the global authoritative evaluation platform Artificial Analysis's July voice rankings.
Speech interaction is becoming a key entry point for AI applications. From intelligent assistants to real-time translation, from meeting minutes to audiobook content generation, the demand for high-quality speech models is increasingly urgent. Alibaba's new generation models aim to solve the challenges of recognition, synthesis, and multi-turn dialogue in complex scenarios.
Specifically, Qwen-Audio-3.0-ASR focuses on speech-to-text, supporting recognition in complex contexts and professional fields; Qwen-Audio-3.0-TTS achieves text-to-speech, supporting multiple languages and dialects, with precise control over emotion, tone, and rhythm; Qwen-Audio-3.0-Realtime has end-to-end speech understanding and dialogue capabilities, able to listen and speak simultaneously, interrupt and follow up at any time, and supports tool invocation.
This achievement is no accident. As early as July this year, in Artificial Analysis's evaluation, the series ranked first in the three tracks of speech recognition (ASR), real-time interaction (RealTime), and speech synthesis (TTS), achieving a 'grand slam.' The evaluation covered multiple indicators such as multilingual capability, noisy environments, and naturalness, demonstrating the models' overall strength.
Currently, the Qwen-Audio-3.0 series has been integrated into products such as the Qianwen App, Qianwen Office, and Qoder, providing users with a more natural speech interaction experience. At the same time, the open API service will accelerate innovation in third-party applications, such as intelligent customer service, voice notes, and accessibility assistance.
Industry observers point out that the competition in speech models is shifting from single capabilities to comprehensive performance. Alibaba's leading position in the three core tracks indicates that it has established a first-mover advantage in end-to-end speech technology. With the evolution of multimodal large models, speech interaction is expected to become the new normal in human-machine collaboration. In the future, how to further reduce latency and improve dialect coverage will be important directions for the iteration of this series.
