亞洲爆紅的 AI 聊天機器人：DeepSeek、Kimi、豆包、Qwen 一次看懂（台灣觀點）

At the beginning of 2025, DeepSeek suddenly gained popularity, making many Taiwanese people realize for the first time that AI is not limited to just a few companies in Silicon Valley. In fact, Asia, particularly China, has produced a batch of its own AI chatbots in the past two years, with user numbers reaching hundreds of millions. This article helps you understand the main players and honestly discusses what you should be aware of.

First, let's talk about the most famous one: DeepSeek

DeepSeek shocked the world because it achieved reasoning capabilities comparable to top models at a relatively low cost, and it is open-source and affordable. For developers, its open-weight model can be self-deployed, which is particularly valuable - you don't have to send your data to someone else's cloud to use a powerful reasoning model.

The king of consumer-end: Doubao and Kimi

If we're talking about what ordinary people use every day, Doubao (produced by ByteDance) is one of the most widely used AI assistants in China, backed by the Douyin ecosystem, and focuses on being easy to use. Kimi (Moonshot AI) has made a name for itself with 'ultra-long text', able to read extremely long files at once, and is particularly strong in long-text summarization and reading academic papers.

Major manufacturers each have their own strengths: Qwen, Wenxin, Zhipu, Baichuan, Tiangong

Alibaba's Qwen Tongyi Qianwen is one of the most popular Chinese models in the open-source community, with comprehensive capabilities; Baidu's Wenxin Yiyin is one of the earliest commercial models; Tsinghua-related Zhipu GLM and Wang Xiaochuan's Baichuan Intelligent are both known for their open-source capabilities; Kunlun Wanwei's Tiangong focuses on AI search. Just in China, the landscape is already fragmented.

Taiwanese people, here are a few things you must know

This section is the most important, please read it carefully.

First, registration and access have barriers. Most Chinese AI services' web/App versions require a Chinese phone number or specific procedures, and overseas registration may not be smooth.

Second, content is affected by local regulations. When it comes to sensitive topics such as politics and history, these services will filter or refuse to answer, which is a product design fact, and you should be prepared when using them.

Third, and most crucially - data storage and privacy. Sending data to these cloud services is equivalent to the data entering China and being subject to local laws and regulations. Therefore, do not put personal sensitive data, company secrets, or customer data into them.

So, how can you use them safely? The answer is: use their open-source models and deploy them locally or on a trusted cloud. DeepSeek, Qwen, Zhipu, and Baichuan all have open-source versions, which allow you to enjoy the model's capabilities without having to hand over your data. For Taiwanese developers, this is the most practical and safest way to access these Chinese AI models.

In a nutshell

The rise of Asian AI is real, and its capabilities are not to be underestimated; but 'being able to use' and 'how to use' are two different things. Using open-source capabilities for yourself and keeping sensitive data in your own hands is the smartest attitude for Taiwanese users. If you want to see Asian AI video and image editing tools, you can continue reading our Asian AI Video Generation War.