Qwen is a family of AI language models made by Alibaba Cloud, also known as Tongyi Qianwen. Many Qwen models have open weights, so developers can download and run them, and Alibaba also offers a Qwen chatbot app. The detail people miss: licences differ between models, so check before commercial use.
Qwen is one of the most important AI model families that many people in the UK and US have never heard of. It comes from Alibaba, it powers a popular chatbot app, and many of its models are open, which means developers can download and run them. If you've seen Qwen listed in a tool's model menu and wondered what it is, this guide is for you.
I'll explain what Qwen is, who makes it, the different types of Qwen models, how you can use them, and the licence and privacy points a business should check first.
What Qwen is
Qwen is a family of AI language models developed by Alibaba Cloud, the cloud computing arm of the Chinese tech company Alibaba. Wikipedia describes it as a family of predominantly open-weights large and small language models. In Chinese it's known as Tongyi Qianwen.
Alibaba launched a beta of Qwen in 2023, and the family has grown quickly since, with new versions and new types of model released regularly.
What open weights means for Qwen
When a model has open weights, its maker publishes the trained model files so anyone can download them and run them on their own hardware. That's very different from assistants you can only use through the maker's own app or service.
Wikipedia notes that many Qwen models are released under the free and open-source Apache License, while some use Alibaba's own Qwen License or a non-commercial research licence. Alibaba has also kept some of its most advanced models proprietary. So 'Qwen is open' is only partly true: it depends on the exact model.
The different types of Qwen models
| Type of Qwen model | What it handles |
|---|---|
| Core Qwen language models | Chat, writing, reasoning and general questions |
| Qwen-VL | Images and text together, using a vision transformer combined with a language model |
| Qwen-Audio | Spoken input and other sounds |
| Coder models | Writing and explaining code |
| Maths models | Mathematical problem solving |
| Omni models | Text, images, audio and video as input |
Wikipedia mentions, for example, an Omni model that accepts text, images, audio and video as input. Version numbers change often, so check Alibaba's or Qwen's own pages for the latest releases.
How you can use Qwen
- The Qwen app: Alibaba offers a Qwen chatbot app, which it has connected to its wider ecosystem of services.
- Run it yourself: open-weight Qwen models can be downloaded and run locally. Ollama, a popular tool for running models on your own computer, lists Qwen among its models.
- Through cloud services: developers can reach Qwen models through Alibaba Cloud and other platforms that host open models.
- Inside other apps: some AI tools let you choose Qwen from their model menu.
Why businesses look at Qwen
For a small business, Qwen usually comes up for one of three reasons:
- Privacy: running an open model on your own computer keeps your prompts on your machine.
- Choice: having another strong model to compare against the big names.
- Building: developers can customise or fine-tune open models for a specific job.
If privacy is your main concern, it's worth comparing running a model locally with the settings available in mainstream assistants; my guide to whether ChatGPT is private covers what to check there.
Qwen licence, privacy and safety checks
Before you use any Qwen model in your business, I'd check these points:
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
- The exact licence. Apache-licensed models are flexible, but research-only licences don't allow commercial use.
- Where your data goes. Running a model locally is very different from using a hosted app or cloud service; read the privacy terms of whichever you use.
- Accuracy. Like every AI model, Qwen can get things wrong, so check facts and numbers that matter.
- Your sector's rules. Some organisations have policies on which AI providers they can use; check yours.
Getting started with Qwen in three simple ways
- Try the Qwen chatbot app to get a feel for how it answers your kind of questions.
- If you have a reasonably powerful computer, install a local model runner and download a smaller Qwen model to test privately.
- If you have a developer, ask them to compare a Qwen model with your current assistant on a real task, such as summarising customer emails.
Whichever route you choose, start with low-risk, non-confidential content, and keep a note of which model and licence you used so you can check later.
Qwen compared with ChatGPT, Claude and Gemini
Qwen's big difference is openness: many of its models can be downloaded and run anywhere, while ChatGPT, Claude and Gemini are mainly used through their makers' own services. For everyday business writing and research, most people stick with a mainstream assistant, and my comparison of Claude and ChatGPT helps you choose between two of them.
If you're curious about trying different models, including open ones, my guide to AI agents that are free to try is a good next step. And if you'd like help deciding which models suit your business and how to use them safely, that's the kind of work I do as an AI consultant.
Frequently asked questions
Who makes Qwen?
Qwen is developed by Alibaba Cloud, the cloud computing arm of Alibaba. In Chinese it's known as Tongyi Qianwen.
Is Qwen open source?
Many Qwen models are released with open weights under the Apache License, but some use more restrictive licences and some advanced models are kept proprietary. Check each model's licence.
Can I run Qwen on my own computer?
Yes, open-weight Qwen models can be run locally. Tools such as Ollama list Qwen among the models they support.
Is Qwen safe for business use?
It depends on how you use it. Check the model's licence, where your data is processed, and your organisation's policies, and always review the output.
What is Qwen-VL?
Qwen-VL is a series of Qwen models that handle images and text together, combining a vision transformer with a language model.
