Tongyi Qianwen (Qwen)

Full-range, open-source, multimodal, and multi-functional

Contact Sales Qwen on Model Studio

Qwen3 Unveilled: Think Deeper, Act Faster

About Qwen

About Qwen

Alibaba Cloud provides Tongyi Qianwen (Qwen) models, a series of large language models (LLMs) and multimodal models (MLLMs), to the open-source community. The latest Qwen3 models adopt hybrid thinking modes ("Thinking" and "Non-Thinking") that allow you to flexibly control reasoning performance, speed, and costs. Lightweight yet powerful, Qwen3 models achieve competitive results in benchmark evaluations of coding, math, general capabilities, etc., when compared to other top-tier models. With support for 119 languages/dialects and Model Context Protocol (MCP), Qwen3 breaks language barriers and enhances Agent capabilities, setting a new benchmark for open-source and proprietary AI models worldwide.

Multimodal models, including Qwen-VL (large language vision model) and Qwen-Audio (large language audio model), support cross-modal processing, while the latest Qwen2.5-Omni enables real-time streaming responses to text, images, audio, and video via its Thinker-Talker architecture.

Leading Performance in Multiple Dimensions

Qwen outperforms other open-source baseline models of similar sizes on a series of benchmark datasets that evaluate natural language understanding, mathematical problem-solving, coding, etc.

Easy and Low-Cost Customization

You can deploy Qwen models with a few clicks in PAI-EAS, and fine-tune them with your data stored on Alibaba Cloud, or external sources, to perform industry or enterprise-specific tasks.

Applications for Generative AI Era

You can leverage Qwen APIs to build generative AI applications for a broad range of scenarios such as writing, image generation, audio analysis, etc. to improve work efficiency in your organization and transform customer experience.

What Qwen Can Do

Analyzing and Understanding
Generating Content
Multimodal Communication
Using Tools (Agentic Capabilities)

Understand Structured Data

Qwen 2.5 understands structured data (such as tables) better. This contributes to extracting insightful information from structured data, helping users perform queries, and generating new datasets. For example, Qwen2.5-72B can provide formatted outputs based on the requirement and input data (table in JSON format).

Identify Objects in Images

Qwen-VL learns and analyzes objects and texts in images and creates new content based on its learning. For example, it can recognize the woman and the dog in the picture and their gestures (high five).

Analyze Images

QVQ can analyze and understand the content that the two images depict in real life, understand the user request, analyze the content, and use CoT to “think” and find the connection between the content in these images.

Understand and Analyze Audio

Qwen-Audio can accept diverse types of audio (such as human speech, natural sounds, instrumental music, and songs) and text as inputs, understand the audio content, and summarize information such as music genres and emotions of the speaker. It can also use tools to edit the audio files.

Understand Multimodal Data

You can use Qwen models (Qwen, Qwen-Audio, and Qwen-VL, etc.) to build a chat assistant that interacts with users intelligently and comprehensively and understands multimodal data, including text, images, audio, and videos.

Generate Long Text

Qwen2.5 significantly improves long text generation, increasing from 1K to over 8K tokens. For example, Qwen2.5-72B can write a report of over 5,000 Chinese characters on the requested subject.

Generate JSON Code

Qwen 2.5 offers improved and more reliable generation of structured outputs. It can generate JSON code step by step with explanations as requested.

Generate Images

Based on text prompts and input images, Qwen-VL can produce high-quality images in various styles and genres (for example, a cartoon-style human portrait) for different industry-specific scenarios.

Generate Videos

Wan excels at generating realistic videos by accurately handling complex movements, enhancing pixel quality, adhering to physical principles, and optimizing the precision of instruction execution.

See, Hear, Talk, Write, Do It All!

Qwen-Omni provides exciting multimodal features, including video and audio interactions, image analysis, text communication, etc. For example, Qwen-Omni can provide real-time guidance on how to complete and improve a painting, identify multiple persons in a video, or describe real-life scenes and provide advice based on request.

Think and Interact with the Environment Using Tools

Qwen3-32B uses tools to fetch information (Repository Stars) from a GitHub page and produce a bar chart to illustrate the information, then organize local desktop files by file type.

What Qwen Can Do

Understand Structured Data

Qwen 2.5 understands structured data (such as tables) better. This contributes to extracting insightful information from structured data, helping users perform queries, and generating new datasets. For example, Qwen2.5-72B can provide formatted outputs based on the requirement and input data (table in JSON format).

Identify Objects in Images

Qwen-VL learns and analyzes objects and texts in images and creates new content based on its learning. For example, it can recognize the woman and the dog in the picture and their gestures (high five).

Analyze Images

QVQ can analyze and understand the content that the two images depict in real life, understand the user request, analyze the content, and use CoT to “think” and find the connection between the content in these images.

Understand and Analyze Audio

Qwen-Audio can accept diverse types of audio (such as human speech, natural sounds, instrumental music, and songs) and text as inputs, understand the audio content, and summarize information such as music genres and emotions of the speaker. It can also use tools to edit the audio files.

Understand Multimodal Data

You can use Qwen models (Qwen, Qwen-Audio, and Qwen-VL, etc.) to build a chat assistant that interacts with users intelligently and comprehensively and understands multimodal data, including text, images, audio, and videos.

Generate Long Text

Qwen2.5 significantly improves long text generation, increasing from 1K to over 8K tokens. For example, Qwen2.5-72B can write a report of over 5,000 Chinese characters on the requested subject.

Generate JSON Code

Qwen 2.5 offers improved and more reliable generation of structured outputs. It can generate JSON code step by step with explanations as requested.

Generate Images

Based on text prompts and input images, Qwen-VL can produce high-quality images in various styles and genres (for example, a cartoon-style human portrait) for different industry-specific scenarios.

Generate Videos

Wan excels at generating realistic videos by accurately handling complex movements, enhancing pixel quality, adhering to physical principles, and optimizing the precision of instruction execution.

See, Hear, Talk, Write, Do It All!

Qwen-Omni provides exciting multimodal features, including video and audio interactions, image analysis, text communication, etc. For example, Qwen-Omni can provide real-time guidance on how to complete and improve a painting, identify multiple persons in a video, or describe real-life scenes and provide advice based on request.

Think and Interact with the Environment Using Tools

Qwen3-32B uses tools to fetch information (Repository Stars) from a GitHub page and produce a bar chart to illustrate the information, then organize local desktop files by file type.

Try Qwen Models on Alibaba Cloud Model Studio

Learn More

Friends of Qwen

icon