What it is
Qwen Chat, which Alibaba now brands Qwen Studio at the same address, chat.qwen.ai, is the free assistant for the Qwen model family. It offers chat, reasoning, deep research, web development, file and image understanding and image generation.
The models behind it moved quickly in 2026: Qwen 3.5 in February, 3.6 in April, 3.7-Max in May and Qwen3.8 in August. Qwen3.8 brought a Max-class model to open release for the first time: Qwen3.8-2.4T-A95B, a mixture-of-experts model with 2.4 trillion parameters and 95 billion active, under a custom Qwen3.8-Max licence, alongside a 27B model under Apache 2.0.
In China, the Qwen App has become an agent that can order food, pay through Alipay, book travel and shop Taobao's catalogue. Those features are not available internationally. Developers pay per token for the same models through Alibaba Cloud Model Studio and QwenCloud.
Our review
Qwen's importance lies in its models, and the chat app is mostly a free demonstration of them. In August 2026 Alibaba released Qwen3.8-2.4T-A95B as open weights, the first time a Max-class Qwen model has been downloadable, plus a 27B version that logged about 7.7 million Hugging Face downloads in the past month. Alibaba runs many of its own published coding benchmarks through Claude Code, a sign of where it expects the models to be used.
As an assistant, Qwen Studio covers the basics well internationally: chat, reasoning, deep research, file and image understanding, and small web apps built from a prompt. In China it goes much further. Since January 2026 the Qwen App can order food, pay in chat with Alipay and book travel, and in May it was connected to Taobao's full catalogue. None of that reaches users elsewhere.
The drawbacks mirror DeepSeek's. Chats are processed by a Chinese company, politically sensitive topics are filtered, and the free app offers no admin console or business data terms. The licence matters too: the flagship open model ships under a custom Qwen3.8-Max licence rather than Apache 2.0, so legal teams should read it before building on it, while the 27B model stays Apache 2.0.
For developers, the API is the better route to the same models. Model Studio lists Qwen3.8-Max at $2 per million input tokens and $6 per million output on its international deployment, below the $2.50 and $7.50 it charges for the older Qwen3.7-Max. A newer model at a lower price is a fair measure of how hard Alibaba is pushing for share.
Where it earns its place
An ML engineer choosing an open model to self-host
Test Qwen3.8 quality before committing GPU budget
Try prompts free in Qwen Studio, then serve Qwen3.8-27B locally; Alibaba's own vLLM example uses four-way tensor parallelism and a 262,144-token context. The 2.4T flagship is far beyond a single GPU server and comes with a custom licence to review.
A startup CTO paying for a frontier model API
Cut model spend on a high-volume agent
Qwen3.8-Max on Model Studio costs $2 per million input tokens and $6 per million output, and QwenCloud accepts OpenAI- and Anthropic-style requests, so switching is mostly configuration. Data then sits with Alibaba Cloud, so check the deployment region before sending customer records.
A graduate student outside China
Get a free second opinion from a strong model on code and papers
Qwen Studio is free with no paid tier and handles drafting, coding questions and document summaries. Keep personal data out, since Alibaba processes it, and expect refusals on politically sensitive research topics.
Before you commit
- Setup
- Sign in at chat.qwen.ai or the mobile apps; open weights download from Hugging Face or ModelScope and run on Transformers, vLLM, SGLang, llama.cpp or MLX.
- Integrations
- QwenCloud and Model Studio APIs accept OpenAI and Anthropic request formats; Qwen Code is an open-source terminal agent tuned for Qwen models.
- Data and privacy
- Chats are processed by Alibaba, and the free app offers no admin controls or business data terms. Model Studio API users choose a deployment scope, such as International (Singapore) or mainland China.
- Cost at scale
- A workload of 100 million input and 20 million output tokens a month on Qwen3.8-Max costs about $320 at list price on the international deployment, before context-caching discounts.
- Limits
- Qwen3.8-2.4T-A95B has a native 262,144-token context, extensible to about 1 million; the app filters politically sensitive topics.
What's good
- Free access to Alibaba's newest Qwen models
- Qwen3.8 flagship released as open weights
- API at $2/$6 per million tokens for Qwen3.8-Max
- Agentic shopping and payments in China
What's not
- Agent features limited to China
- Chinese data handling and content filtering
- Flagship open model under a custom licence, not Apache 2.0
- Confusing naming across Qwen Chat, Qwen Studio, Qwen App and QwenCloud
Best for
Developers and curious users who want to try Qwen models for free before using them via API or self-hosting.
Alternatives
Qwen Chat: common questions
- Is Qwen Chat worth it?
- A free, capable window onto Alibaba's Qwen models, but outside China the models matter far more than the app, which adds Chinese data and content rules without much in return. We rate it 3.6 out of 5.
- How much does Qwen Chat cost?
- Free on the web and in apps. API: Qwen3.8-Max costs $2 per million input tokens and $6 per million output on Model Studio's international deployment, with a 1-million-token free quota for new users in Singapore.
- What are the downsides of Qwen Chat?
- Agent features limited to China; chinese data handling and content filtering; flagship open model under a custom licence, not Apache 2.0; confusing naming across Qwen Chat, Qwen Studio, Qwen App and QwenCloud.
- Who is Qwen Chat best for?
- Developers and curious users who want to try Qwen models for free before using them via API or self-hosting.
- How much does Qwen Chat cost for a team or in production?
- A workload of 100 million input and 20 million output tokens a month on Qwen3.8-Max costs about $320 at list price on the international deployment, before context-caching discounts.
- Is Qwen Chat safe for company data?
- Chats are processed by Alibaba, and the free app offers no admin controls or business data terms. Model Studio API users choose a deployment scope, such as International (Singapore) or mainland China.
- What are the best Qwen Chat alternatives?
- The closest alternatives we have reviewed are DeepSeek, Kimi and Mistral Vibe (formerly Le Chat).
Answers checked Oct 22, 2026.