What Qwen is
Qwen, also called Tongyi Qianwen, is a family of mostly open-weight language models made by Alibaba Cloud. Alibaba launched a beta in April 2023 and opened it for public use in September 2023 after regulatory clearance, according to Wikipedia. Open-weight means the model files are published for anyone to download.
The free chat lives at chat.qwen.ai. Qwen's own GitHub page now calls it Qwen Studio and describes it as free to use and open to all, and older tutorials call it Qwen Chat. Wikipedia says the Qwen app had 234 million users by May 2026.
On 6 November 2026 the newest family is Qwen3.8. Alibaba released the cloud model Qwen3.8-Max on 3 August 2026, published the weights of a 2.4 trillion parameter version on 12 August, and released the smaller Qwen3.8-27B on 14 August under the Apache 2.0 license. Its model card says it understands images and video and thinks before it answers by default, a switch you can turn off.
What Qwen is good for
- Everyday questions and drafts: answers, emails and translations. The Qwen3.5 notes on GitHub list support for 201 languages and dialects.
- Reading images and documents: the Qwen3.8-27B card says it handles images and video, from diagrams and documents to hour-long clips. A 2025 tutorial by Alexey Makarov also covers text extraction from screenshots and photos.
- Deep research and web pages: Qwen's GitHub page says its official sites add deep research, web development and adaptive tool use on top of the model.
- Projects for your files: Qwen's own video shows a project that holds your documents, custom instructions and a memory limited to that project.
- Coding: Qwen Code is an open-source AI agent for the terminal, optimized for Qwen models, and Alibaba says Qwen3.8 improves coding and long, multi-step tasks.
- Your own copy: the weights are on Hugging Face and ModelScope, and the API on Qwen Cloud follows the OpenAI and Anthropic formats.
When to use Qwen, and when to pick another
| Your job | Use Qwen? | Or try |
|---|---|---|
| A free everyday chatbot that can also read images and files | Yes. Qwen Studio is free to use | ChatGPT, Claude or Gemini |
| Running an open model on your own computer | Yes, with a small size. Qwen is a common first pick in Ollama | Ministral 3 or Gemma, which the local AI lesson also uses |
| Private files, client data or anything under a workplace rule | Not in the hosted chat, which runs on Alibaba's servers | A small Qwen model through Ollama or LM Studio |
| A research report with sources from the live web | Partly. Deep research is on the official site | Perplexity, which is built for cited answers |
| Coding in a terminal | Yes, with Qwen Code | Claude Code or GitHub Copilot |
| Fine-tuning a model of your own | Yes. Wikipedia calls Qwen a common starting point for fine-tuned models | The Hugging Face lesson shows where to start |
How to use Qwen, step by step
These steps follow the web chat. Qwen renames and adds features often, so follow the names on your screen where they differ from the names here.
- Open Qwen. Go to chat.qwen.ai. Qwen also has a mobile app and, in a July 2025 video from its own channel, a desktop app.
- Sign in. Create an account or log in when the page asks. The chat is free to use.
- Ask a question. Type in the message box. Say what the answer is for, such as a reply to a customer or a study guide.
- Pick the model. Open the model menu and choose the newest model it offers. The Qwen3.8 family arrived in August 2026.
- Add what the question needs. Attach an image or a file with the attachment control, and switch on web search or deep research if the box shows them.
- Make a project for repeat work. Group the files and your standing instructions in one project, as Qwen's own project video shows.
- Try the other outputs. Ask for a small web page, a set of slides or a logo idea, and read the way the answer is laid out.
- Check what matters. For a figure, a date or a legal or health point, open a source you trust before you rely on it.
Type your question in the Message Qwen box in the middle of the page, then press Enter to send it.
Press the model name at the top left, Qwen3.8-27B in this picture, to open the model menu and pick the newest model on offer.
Press the + to attach an image or a file. Switch on Search or Deep Research beside it when the answer needs the web.
Press Projects in the left column to keep the files and standing instructions for one job together.
AI-made illustrations of Qwen's screens, drawn on 6 November 2026 to match these steps. They are not screenshots, so the real app may look different: follow what you see.
The tutorials at the top of this page show these steps on screen. Some were made before the rename, so they say Qwen Chat.
An example to try
Use a receipt or table where you already know the right total. That way you can judge how well the model reads the image.
Attach a photo of a receipt, or a screenshot of a table you know well.
Read out every line and its amount as a table, then add the total.
Say which entries you could not read clearly.- Compare each line with your own copy. Note any digit it misread, because small figures are where reading errors hide.
- Add the amounts yourself and compare with its total. A wrong total with correct lines is a calculation slip, not a reading slip.
- Ask it to turn the table into a short summary for a spreadsheet, and check the numbers once more.
Tips for better results
- Crop the image to the part that matters, and use a sharp photo in good light.
- Name the form you want: a table, three bullets or a short email.
- Keep thinking on for hard problems, and switch it off for quick rewrites if the model offers the choice.
Privacy, licenses and running Qwen yourself
The hosted chat runs on Alibaba's servers, so what you type goes there. Alibaba is a Chinese company, so read its privacy policy, follow your employer's rules and keep client files and personal details out of the chat.
Not every Qwen model is open. Wikipedia says some top models, such as Qwen3.7-Max and the Plus models, are proprietary and reach you through the chat sites and Alibaba Cloud. The open ones carry the Apache 2.0 license, the Qwen License or a research license. The 2.4 trillion parameter weights ask large service providers to get a commercial license, so check the license on the model page.
To run Qwen at home, pick a small size. Ollama's own account announced the Qwen3.5 small models as ollama run lines, from qwen3.5:0.8b up to qwen3.5:9b. Qwen's Qwen3 page warns that Ollama's names can differ from Qwen's own, so read the tag list before you pull. A 2026 tutorial by RepoChad says Qwen3.8-27B takes roughly 54 GB in full precision and shows setups for graphics cards from 8 GB to 32 GB.
Learn more about Qwen
- 2Which AI model should I use for my task?Which AI model should I use? Pick the assistant by your task and where your files live, then a fast or deep-thinking model, and test two on your own work.
- 1Ollama tutorial: run AI models on your own computerAn Ollama tutorial for Windows, macOS and Linux: install it, download and chat with a first local model, then list, stop and remove models and move their files.
- 1Hugging Face tutorial: find, check and download open modelsA Hugging Face tutorial for Python users: find a model on the Hub, read its card and license, request a gated model and download it with the hf CLI.
Frequently asked questions
What is Qwen?
It is a family of mostly open-weight language models from Alibaba Cloud, also called Tongyi Qianwen. It began as a beta in April 2023 and opened to the public in September 2023.
Is Qwen free?
Qwen's GitHub page says Qwen Studio at chat.qwen.ai is free to use. The open models are free to download, and the paid API on Qwen Cloud is billed by use, so read its pricing page before you build on it.
What is Qwen Studio?
It is the name Qwen's GitHub page now uses for the chat at chat.qwen.ai, which older tutorials call Qwen Chat. It adds deep research, web development and adaptive tool use on top of the models.
Is Qwen open source?
Many of its models are open-weight, and several carry the Apache 2.0 license, such as Qwen3.8-27B. Some top models are proprietary, and the biggest open weights have a license that asks large providers for a commercial deal.
Can I run Qwen on my own computer?
Yes, a small one. Ollama's own account lists Qwen3.5 sizes from 0.8b to 9b. The 27B model needs a graphics card with 8 GB or more of video memory when quantized, according to a 2026 tutorial.
Is Qwen safe to use?
Keep private data out of the hosted chat, because it runs on Alibaba's servers. For sensitive files, run a small open Qwen model on your own computer instead.
How is Qwen different from DeepSeek?
Both are Chinese makers with open-weight models. Qwen comes from Alibaba Cloud and has many sizes and licenses, and its chat reads images and files. Try both on your own task, as the model lesson explains.
Sources
- Wikipedia: Qwen https://en.wikipedia.org/wiki/Qwen
- GitHub: QwenLM/Qwen3.8 README https://github.com/QwenLM/Qwen3.8
- GitHub: QwenLM/Qwen3 README https://github.com/QwenLM/Qwen3
- Hugging Face: Qwen/Qwen3.8-27B https://huggingface.co/Qwen/Qwen3.8-27B
- Ollama on X: the Qwen 3.5 small model series is now available https://x.com/ollama/status/2028510184788926567
- YouTube: Qwen, How to Use AI to Manage Your Health and Insurance, Qwen Chat Project Tutorial https://www.youtube.com/watch?v=M_S5COpcixk
- YouTube: Unreasonable AI, How to Use Qwen AI from Alibaba for Beginners https://www.youtube.com/watch?v=hc4Xgj6NUcc
- YouTube: Learn AI Tools, How to Get Started Using Qwen AI https://www.youtube.com/watch?v=36zn_Ts4cd0
- YouTube: RepoChad, Run Qwen3.8-27B on ANY GPU https://www.youtube.com/watch?v=0xUxO_9zqTU
- YouTube: Qwen, Qwen Chat for Desktop is here https://www.youtube.com/watch?v=V4MCmMQo_B4