Qwen Studio is Alibaba's official chat assistant built on the Qwen family of neural networks: the site chat.qwen.ai and a phone app. The service used to be called Qwen Chat. The chat is free to use, while for code and your own services Qwen has a separate paid API. Below: which addresses are real, what you can download, how access from Russia looks and how to connect Qwen models by key.
What Qwen Studio is and how it differs from Qwen Chat
Qwen Studio is a product for people: an interface, not an API for programs. In the README of the Qwen3.8 repository, the Qwen team describes it as an assistant for everyone, free and open to any user. The same README says the official sites have extra features: deep research, web development and adaptive tool selection.
Qwen Chat was renamed Qwen Studio in spring 2026. Third-party sources give the date as April 10, while the official materials we managed to open contain no announcement. According to those same sources, user data and settings were kept. So an instruction about "Qwen Chat" that you came across describes the same service.
Alibaba has several more products around the chat, and the confusion between them explains half of all "qwen official site" searches:
| Product | What it is | Address |
|---|---|---|
| Qwen Studio | A chat and app for people | chat.qwen.ai |
| QwenCloud | A paid API for developers | qwencloud.com, docs at docs.qwencloud.com |
| Qwen Code | A terminal coding agent | github.com/QwenLM/qwen-code |
| Open weights | Models to run yourself | huggingface.co/Qwen, modelscope.cn |
| Qoder and QwenWork | Platforms with Qwen models inside | qoder.com, qwenwork.cn |
The official Qwen site: which addresses are real
The Qwen3.8 README, the team's own repository, lists the chat, the API, Qoder, QwenWork and the places where the weights are published. The qwen.ai page links to the blog and to Qwen Code, and the agent's repository sits in the QwenLM organization on GitHub. API documentation opens from a link on the QwenCloud home page. chat.qwen.ai is assembled dynamically, so we did not read its contents directly and rely on the README.
Search for "qwen download" returns both APK mirrors and lookalike sites. Three rules cut out almost everything unwanted:
- The domain must match an address from the README: chat.qwen.ai, qwen.ai, qwencloud.com.
- Get the Android app from Google Play, where the publisher is Alibaba Cloud.
- Enter an API key only on qwencloud.com and in trusted clients. A "chat" page never asks for a key.
Downloading Qwen: the app, the weights or the CLI
People mean three different things by "download".
The chat app. Google Play has Qwen Studio by Alibaba Cloud, package ai.qwenlm.chat.android, and its description opens with the words "your ultimate personal AI assistant". We did not confirm iOS or desktop versions in the official sources available to us: the downloads section on qwen.ai is served dynamically. Check it yourself.
A model on your own computer. The open weights are published on Hugging Face and ModelScope. The shortest route is through Ollama:
ollama run qwen3.8
The command downloads the 27B version, about 18 GB. A memory estimate and a breakdown of quantization are in a separate article, Run Qwen 3.8 27B locally.
A coding agent. Qwen Code installs from npm and needs Node.js 22 or newer:
npm install -g @qwen-code/qwen-code@latest
We covered installation on Windows, macOS and Linux, and the comparison with Claude Code, in the article about Qwen Code.
What Qwen Studio can do and what is free in it
The Qwen3.8 README lists three features of the "native" mode on the official sites: deep research, web development and adaptive tool selection. Which models, and with what limits, are open to a free user depends on the account and may change. We found no "feature versus plan" table from Alibaba, and no paid subscription for the chat either.
A practical conclusion follows. For one-off questions, drafts and reading a document, the chat is enough. If you need a repeatable result, cost tracking or calls from your own code, the chat will not do: use the API.
Which Qwen models exist as of October 29, 2026
"Qwen model" with no qualifier says little: Alibaba has several lineups, and access to them works differently. The table follows the Qwen3.8 README, the Hugging Face cards and the qwen3.8-max page on QwenCloud.
| Model | How to get it | What is known |
|---|---|---|
| Qwen3.8-Max | API, ID qwen3.8-max | MoE with 2.4 trillion parameters, 1 million token context, $2 and $6 per 1M tokens on QwenCloud |
| Qwen3.8-2.4T-A95B | Open weights, August 12, 2026 | 95 billion active parameters, text only, a separate qwen3.8-max license |
| Qwen3.8-27B | Open weights, August 14, 2026 | Dense model, Apache 2.0, text, image and video input, 262,144 tokens |
| Qwen3.8-Flash | API, ID qwen3.8-flash | 1 million token context, $0.15 and $0.47 per 1M tokens on QwenCloud |
| Qwen3.8-Flash-Next | Open weights, August 26, 2026 | 125 billion main parameters and 51 billion N-gram embeddings, 6 billion active |
| Qwen3.8-Omni-Flash | API, ID qwen3.8-omni-flash | Text, image, audio and video input, 1 million token context |
For more on the lineup and the vendor's prices, see the Qwen 3.8 review. A task-by-task comparison with Claude is in the article Claude vs Qwen.
Does Qwen Studio work from Russia
We found no official list of countries where the chat is closed, either in the README or on other Qwen pages we could open. The README says "open to all" and is silent about countries. Whether the site opens and registration goes through depends on your network and provider. Checking takes a minute: open chat.qwen.ai with no extra setup and try to sign in.
The API is a different story. According to the documentation, the QwenCloud free quota is usually 1,000,000 tokens per model, applies only in the Singapore region and is granted after you fill in payment details. The documentation does not say which cards the payment form accepts. Keys are tied to a region, and the address depends on it, for example ap-southeast-1.maas.qwencloudapi.com for Singapore.
A summary by task:
| What you need | Route | What to keep in mind |
|---|---|---|
| Talk to Qwen | chat.qwen.ai or the app | Free, check availability from your location |
| Call a model from code | QwenCloud or a gateway | The vendor needs an account with payment details |
| Run it yourself | Open weights | A GPU or a Mac with 24 GB of memory |
| A coding agent | Qwen Code, OpenCode | The model is connected separately |
Qwen through codexkey: key, base URL and model
Let us be plain: codexkey does not sell subscriptions and does not open the Qwen Studio web chat. We provide API access: one sk_cdx_... key, payment by a Russian card, SBP (the Russian Faster Payments System) or crypto, with no Alibaba account and no VPN. The address for OpenAI-compatible clients is https://api.codexkey.online/v1. The current list of IDs comes from GET https://api.codexkey.online/v1/models, and it is also on the models page.
The codexkey price list as of October 29, 2026, prices per 1M tokens; fresh numbers are on the models page:
| ID | Input | Output | Context | Vendor price (QwenCloud) |
|---|---|---|---|---|
qwen3.8-max | $0.28 | $0.86 | 1000K | $2 and $6 |
qwen3.8-flash | $0.10 | $0.28 | 1000K | $0.15 and $0.47 |
qwen3.8-omni-flash | $0.04 | $0.12 | 1000K | see the model page on QwenCloud |
qwen3.8-flash-free | $0 | $0 | 1000K | none |
The vendor has caching and batch processing with separate prices, so we compare only the base rows. For the terms and limits of the free qwen3.8-flash-free, see the model page: the catalog has no limit figures, and we will not invent any. Free DeepSeek models are covered in DeepSeek for free: limits and status.
A minimal call through the OpenAI SDK:
from openai import OpenAI
client = OpenAI(
base_url="https://api.codexkey.online/v1",
api_key="sk_cdx_YOUR_KEY",
)
reply = client.chat.completions.create(
model="qwen3.8-max",
messages=[{"role": "user", "content": "How does Qwen Studio differ from QwenCloud? Three points."}],
)
print(reply.choices[0].message.content)
To connect Qwen to a terminal agent, take the config from the OpenCode guide. It describes two codexkey providers, and a Qwen model is added to the codexkey block the same way as deepseek-v4.1-flash. The limit block with the context size must not be removed. A step-by-step OpenCode installation will be in the article OpenCode: installation and models. Making the first call and creating a key are described in the quickstart.
FAQ
Where is the official Qwen site?
The chat runs at chat.qwen.ai, the blog and the Qwen Code page are on qwen.ai, the paid API is at qwencloud.com, and the weights are at huggingface.co/Qwen and modelscope.cn. These addresses are listed in the README of the QwenLM/Qwen3.8 repository. Check any other sites that a search for "qwen official site" returns against this list.
Is Qwen Studio free?
The Qwen3.8 README calls the chat free and open to everyone. We found no ChatGPT Plus-style subscription for it. Alibaba publishes no numeric message limits, so an exact figure cannot be given: the terms are visible in the interface itself. The API stays paid.
How do I download Qwen to a phone or computer?
On Android: the Qwen Studio app by Alibaba Cloud on Google Play. On a computer the chat works in the browser, and models are run locally from the open weights, for example with ollama run qwen3.8. We did not confirm iOS or desktop installers in the official sources available to us.
Does Qwen work from Russia without a VPN?
We found no official ban for Russia, but we have no confirmation either that the site and registration are open to everyone. Check chat.qwen.ai from your own network. The API route through codexkey does not depend on whether Alibaba's site is reachable: requests go to api.codexkey.online.
How does Qwen Studio differ from Qwen Code and QwenCloud?
Qwen Studio is a chat for people. Qwen Code is a terminal agent that edits a project's code. QwenCloud is a paid API through which programs call models. All three run on Qwen models, but they connect and bill differently.
Can I use Qwen Studio itself through codexkey?
No. codexkey gives only access to models by API: a key, a base URL and IDs such as qwen3.8-max. We do not sell the web chat, the app or Alibaba subscriptions. If you need the Qwen Studio interface itself, go to chat.qwen.ai directly.