Qwen AI
Alibaba assistant and open-weight model family, free to chat with in the browser and available as an API.
Quick facts
| Pricing | Free |
|---|---|
| Free tier | Yes — free chat access with usage limits |
| Platforms | Web, iOS, Android, API |
| Developer | Alibaba |
| Website | https://chat.qwen.ai/ |
What Qwen AI is
Qwen is Alibaba model family, and the consumer assistant is its most visible part. The chat handles writing, translation, coding and document questions, with strong Chinese-language coverage and reasonable English. What distinguishes Qwen from most consumer assistants is that many model sizes are published with open weights, so developers can run them locally or fine-tune them, and there are specialised variants for coding, vision and audio. The website offers free chat access; API access to hosted models is billed per token. The limits are documentation and availability: English documentation is thinner than for Western models, some features are region-gated, and the consumer interface changes frequently. For teams with Chinese-language workloads it is often the strongest option that can also be self-hosted. A long-context variant handles book-length inputs, and a coding variant is distributed separately for IDE use. The chat also accepts image and document uploads for questions about a file.
Best for
- Chinese-language writing and translation
- Developers who want an open-weight model they can host
- Cost-sensitive workloads that need a capable model
Not ideal for
- Teams that need extensive English documentation
- Regions where access is restricted
Pros
- Open weights for several model sizes
- Strong Chinese-language performance
- Free chat access and low-cost API
Cons
- English documentation is thinner than rivals
- Some features are region-gated
- Interface and model lineup change frequently
Pricing in detail
Chat access on the website is free with usage limits. Hosted API access is billed per token, with rates below most Western frontier models and separate pricing for larger variants; open weights can be self-hosted at no licence cost. Because self-hosting shifts cost to hardware and operations, the cheaper option depends on whether you already run GPUs.