Qwen's answers shouldn't end in the chat.

The name

Qwen is short for its official Chinese name, Tongyi Qianwen. It names the chat AI and also the whole family of large language models (LLMs) that handle text, images, audio and code.

Where it comes from

Qwen is a Chinese AI, developed by the Qwen team at Alibaba Group. The international Qwen Chat (chat.qwen.ai) is provided by the Singapore company Alibaba Cloud (Singapore), and its privacy policy says input is processed in Singapore. The API is offered through Alibaba Cloud Model Studio, where you can choose from several regions, including Tokyo.

Three ways to use it

• Qwen Chat: talk to it for free in a browser or mobile app. Good for personal research and translation.
• API: build it into your own systems and tools through Alibaba Cloud Model Studio. You pay for what you use.
• Open weights: run models (the weight files) published on Hugging Face or ModelScope on your own PC or server. Your data stays in-house.

What Qwen can do

• Chat: answering questions, writing, translating and summarizing. Web search and Deep Research too.
• Images and video: understanding images and video, and generating and editing images.
• Audio: models for speech synthesis and real-time translation.
• Code: coding models, and Qwen Code, which runs in the terminal.

Many of its small and mid-size models are open weights you can run on your own PC or server. That is the big difference from ChatGPT and Claude.

1. Open chat.qwen.ai. It works in the browser. On a phone, there's a "Qwen" app for iOS and Android.
2. Log in. Create an account and log in.
3. Pick a model. Switch models with the model picker. If unsure, keep the default.
4. Start talking. Ask in English or many other languages. Attach files or images and ask it to "summarize this" or "make a table".
5. Switch features. Choose web search, Deep Research, image generation, Web Dev (quick web page prototypes) and more.

Qwen Chat is free. As of October 2026, we found no paid personal plan for Qwen Chat like ChatGPT Plus.

Tips for work

• State the role and format first: say who it's for and what shape you want up front, like "a reply to a supplier, formal, three paragraphs". You'll edit less.
• Check translations against the original: for amounts, dates, model numbers and names, compare with the source rather than reading only the translation.
• Give long documents as files: attaching a PDF or document keeps tables and headings intact better than pasting.
• Separate conversations: start a new chat for each project so earlier topics don't skew the answers.

Turning those answers into who does what by when, and following up, happens outside Qwen Chat. If you're doing that yourself every time, see how SIMY connects.

Scroll the table sideways →

Next generation: at the Apsara Conference on September 22, 2026, Alibaba announced that Qwen 4 is in training (no release date announced). Model names change quickly, so it's enough to remember the roles: "top tier", "mid-size for local use" and "built for speed".

How to use Qwen Image Edit

Attach an image in Qwen Chat and describe the change, like "make the background white" or "translate the sign into English". The open-weight version is often built into image tools such as ComfyUI.

• Where to start: get an API key from Alibaba Cloud Model Studio (DashScope). The API is OpenAI-compatible, so existing tools work by switching the endpoint.
• Regions: for Qwen3.8-Max you can choose Beijing, Singapore, Frankfurt, Virginia, Tokyo or Hong Kong. Available regions differ by model, and this choice decides where your data is processed.
• Pricing: pay per input and output token. As of October 2026, qwen3.8-max is officially listed at $1.65 input and $4.951 output per million tokens (Tokyo and others; Singapore is $2.00 and $6.00). Rates also vary by model, thinking mode and caching, so check the official pricing page for current prices (and for how tax applies).
• Coding Plan: a flat-rate plan for using Qwen and other models from coding tools such as Claude Code and Qwen Code. As of October 2026 only Pro ($50 a month) is offered, with request limits per 5 hours, week and month. See the official description.

Getting started with the API

1. Create an Alibaba Cloud account. Open the Model Studio console and choose your region.
2. Create an API key. Keep it in an environment variable, never directly in code or shared documents.
3. Point to the OpenAI-compatible endpoint. With the OpenAI SDK or existing tools, just change the base URL and model name.
4. Start small and watch usage. Thinking mode and long context use many tokens. Check daily usage in the console before scaling up.

How SIMY relates: SIMY cannot connect to a Qwen API key at this time. SIMY's cloud runs use your ChatGPT (Codex) account.

• Ollama: install it and type ollama run qwen3.8 in a terminal to run the 27B model (about an 18 GB download, takes text and image input). As a rule of thumb, you need a PC with plenty of memory.
• LM Studio: search for and download models from the app. Good if you'd rather avoid the command line.
• Others: llama.cpp, Jan, vLLM, SGLang and more. With vLLM or SGLang you can stand up an internal OpenAI-compatible server (an API called the same way as OpenAI's).

License differences

"Qwen is open source" is too broad. Many small and mid-size models such as Qwen3.8-27B use Apache 2.0, which makes commercial use easy. The largest, Qwen3.8-2.4T-A95B, has a custom license, and API-only models such as Max have no published weights.

Work that suits local models

Local models shine at summarizing internal documents you don't want to send out, working offline, and producing routine drafts in volume every day. Small models are less capable than large ones, though, so use them to make first drafts rather than final versions. Pick one use first, and it's easier to judge what model size is enough.

Memory requirements, choosing a quantization and model size guidelines are covered in How to run Qwen locally.

Qwen Code is an open-source (Apache 2.0) coding agent from the Qwen team. It runs in the terminal, reads and edits code, and runs commands.

• Install: with Node.js 22 or later, run npm i -g @qwen-code/qwen-code@latest. On a Mac you can also use Homebrew.
• Models it connects to: Qwen, plus the OpenAI, Anthropic and Gemini APIs, and local models served by Ollama or vLLM.
• Who it's for: developers who want a Claude Code or Codex style workflow with Qwen models or their own servers.

First steps

1. Run qwen in the folder you want to work in. First, choose the model and how to authenticate.
2. Describe what you want in plain language. Start from the goal, like "explain how this folder is organized" or "find why this test fails".
3. Review changes before accepting. Approve file edits and commands after reading them. Starting with small fixes is safer.

After you write code with Qwen Code, SIMY picks up the fixes and assignments agreed in review meetings and lists them.

The Qwen3 launch claimed support for 119 languages and dialects (the 3.8 model cards don't state a number). Third-party evaluations have also reported strong Japanese performance among models of similar size.

• Strengths: translating between Chinese, English and other languages, summarizing long documents, drafting emails and reports, organizing tables.
• Watch out for: local proper nouns, fine distinctions of formality, and statements about local laws and business customs. Check them before use.
• With small models: small local models can sound unnatural in long passages. Finish important text with a larger model to be safe.

When working with suppliers in China, you can have Qwen translate the meeting summary and leave the owners, deadlines and reply draft to SIMY. The one-meeting comparison above is an example.

Scroll the table sideways →

• Check company rules first: some companies set which AI services you may use and what you may enter.
• Keep confidential data out by default: with any chat AI, think before entering customer data or unpublished figures.
• Verify the answers: check translated numbers, proper nouns, and statements about laws or specs against the original before use.

Qwen Chat data handling is based on the Qwen privacy policy as of October 1, 2026. How SIMY handles data is published on our Security page and Privacy Policy.

Scroll the table sideways →

How to choose

• Keep data in-house: open-weight models you can run yourself, such as Qwen3.8-27B.
• Run high volumes at low API cost: the Qwen or DeepSeek API. Compare regions and rates.
• Value add-on features and familiarity at work: ChatGPT or Claude. If company rules decide, follow them.

Using them together: any AI, Qwen included, can produce the answers and drafts. If you're tracking owners, deadlines, replies and updates yourself every time, that's where SIMY comes in. For how Chinese AI models differ overall, see Chinese AI models (LLMs) compared: DeepSeek, Kimi, GLM and more. For ChatGPT, see the ChatGPT guide.