此语言仍有 24 条内容或界面任务待翻译;未完成部分显示中文原文。
AI Home / Model Library / Qwen Audio (Qwen-Audio)

Qwen Audio (Qwen-Audio)

AI 辅助整理
阿里云音频转录内容摘要情感分析音频事件检测语音聊天音频问答Enter Reference Price · Official Prices Prevail

Qwen Audio is a large-scale audio language model developed by Alibaba Cloud. It supports understanding of multiple audio types, including speech from speakers, natural sounds, music, and singing. Its core capabilities include audio transcription, content summarization, sentiment analysis, audio event detection, and voice chat. Currently, qwen-audio-turbo offers two versions — a stable version and a latest version — supports API calls only in the Beijing region, and provides only a free trial quota; once exhausted, paid calls are not available. It is suitable for validating audio understanding capabilities and running small-scale trials; for production-grade audio understanding, the official recommendation is to use Qwen Omni instead, which supports paid usage and omni-modal capabilities.

Context tokens8,000
Input / Million tokensPending Verification
Output / Million tokensPending Verification
Versionqwen-audio-turbo(稳定版)、qwen-audio-turbo-latest(最新版)
Deployment云端 API:仅支持中国内地部署模式,接入点与数据存储位于北京地域;仅支持通过 DashScope API 调用,暂不支持控制台在线体验

Capabilities & Use Cases

音频转录内容摘要情感分析音频事件检测语音聊天音频问答

Model positioning: Qwen Audio is an audio language model that can recognize speech content and analyze a speaker's gender, age group, accent, emotion, and intent. It can also identify natural sounds such as car horns, thunder, and breaking glass, and analyze instruments, rhythm, tonality, and style in music. Versions and specs: qwen-audio-turbo (stable) offers a context of 8000 tokens, max input 6000 tokens, and max output 1500 tokens; qwen-audio-turbo-latest offers a context of 8192 tokens, max input 6144 tokens, and max output 2048 tokens. Audio is converted at 25 tokens per second, with audio shorter than 1 seconds counted as 25 tokens. The free quota is 100,000 tokens and is valid for 90 days after activating Alibaba Cloud Bailian. Access and invocation: A Beijing-region API Key is required. Calls are supported only via API (DashScope SDK or HTTP); the console's online trial is not supported. Audio can be provided in three ways: public URL, Base64 encoding, or local file path; the local path option is supported only by the DashScope Python and Java SDKs. Input limits: Audio passed via public URL or local path must not exceed 10 MB, and Base64-encoded strings must not exceed 10 MB; audio duration must not exceed 30 seconds, and only the first 30 seconds are processed if exceeded. Supported formats include AMR, WAV (GSM_MS), WAV (PCM), 3GP, 3GPP, AAC, and MP3; supported languages include Chinese, English, Cantonese, French, Italian, Spanish, German, and Japanese. Selection tip: The model has no paid option and is not suitable for direct production use. For production-grade audio understanding or omni-modal capabilities, the official recommendation is to migrate to Qwen Omni, which supports paid usage and covers both the international (Singapore) and mainland China (Beijing) regions.

Capability Comparison

No verified item-by-item scores available; capability radar and overall ranking will not be displayed for now.

Data & Sources

Data Provider
阿里云
On-site profile update
2026-11-06 04:36:05

On-site update time is not equivalent to the model release date. Official or platform benchmark scores are not displayed when test sets, dates, or original results are not provided.

Version History

qwen-audio-turbo(稳定版)、qwen-audio-turbo-latest(最新版)Current Entry Version

Historical versions, release and discontinuation dates to be verified.

Access & Usage

Please confirm the API address, model identifier, authentication, permissions, and billing rules according to the official documentation.

Open Official Documentation

Platform Tested

No verified model test reports available. Model capability tags do not indicate completed platform testing.
Product Selection Assistant

让需求变成可修改的筛选条件

依据已录入参数生成条件,结果可查看和修改。帧率与性能需核对实测,不能仅凭算力数值保证。