Chinese AI is putting pressure on the West

DeepSeek R1 - Chinese AI with powerful reasoning capabilities

Developed by Hangzhou-based startup DeepSeek, DeepSeek R1 is considered the best AI reasoning model in the country with 671 billion parameters. In independent tests, this Chinese AI model has surpassed many leading models such as OpenAI's GPT-4o and Llama 3.1 (Meta) in terms of logical thinking, programming and mathematical ability.

Chinese AI model DeepSeek surprises the world when it announces training cost of only 5.6 million USD

R1 is not only known for its performance but also for its extremely low training costs, which have "stunned" global experts. Accordingly, DeepSeek only costs about 5.6 million USD to train thanks to algorithm and hardware optimization. This is an important factor that helps China maintain its strong AI development capabilities despite the limitations of the chip embargo from the West. Although there are reports that the actual amount is 1.6 billion USD, this cost is still much lower than the investment that Western competitors pour into artificial intelligence.

DeepSeek Hit by Cyberattack After Shaking Tech Stocks

Alibaba Qwen - High Performance, Data Optimization

Alibaba Qwen, especially the Qwen 2.5-Max version launched on the first day of the Lunar New Year and most recently the QwQ-32B introduced in early March 2025, is attracting attention with its superior reasoning ability despite being much smaller in scale than the AI ​​models currently on the market. The QwQ-32B variant with 32 billion parameters but has the same reasoning ability as DeepSeek R1 - a model 20 times larger (671 billion parameters).

Another advantage of Qwen is that it is released as open source, which has helped it quickly become the most popular Chinese AI model on the Hugging Face platform. Its fast processing and data optimization capabilities have helped Alibaba Qwen gain the trust of many domestic businesses in commercial AI applications.

Qwen's superior performance but resource-saving helped Alibaba attract a lot of attention from investors, causing its stock price to surge after the new model was announced. In contrast, this boom pushed NVIDIA's stock down.

Tencent Yuanbao - Chatbot with billion users takes advantage of its own ecosystem

China is not only developing powerful AI models, it is also pushing for practical applications. Tencent Yuanbao, an AI assistant integrated into WeChat, is a typical example.

Launched in May 2024, Yuanbao quickly became the most popular AI app in China, even surpassing DeepSeek in terms of downloads. This was largely due to Tencent integrating Yuanbao into the WeChat ecosystem, giving it an advantage that its Western competitors could not compete with.

Yuanbao leverages WeChat ecosystem and quickly achieves success

Yuanbao leverages WeChat’s vast trove of data to provide accurate, contextual responses, allowing users to look up information, make personal plans, and even make financial transactions within the app. This approach has helped Tencent attract more than 1.3 billion users, making Yuanbao one of the most influential Chinese AI chatbots today.

Manus - The First Fully Automated AI Agent

Developed by startup Monica, Manus is one of the world's first AI agents that can act on its own without constant human guidance. Unlike chatbots that simply respond, Manus can complete tasks independently. For example, when tasked with finding information about an apartment, Manus not only aggregates data but also proactively analyzes price trends, environmental conditions, and recommends suitable listings.

What’s remarkable is that Manus is capable of performing multiple tasks in a row. This Chinese AI can even organize resumes, analyze stocks, and design websites from a single user input. This is a level of AI automation that models like OpenAI’s GPT-4o or even its “compatriot” DeepSeek have not been able to achieve.

Source: https://thanhnien.vn/nhung-ai-trung-quoc-dang-gay-ap-luc-cho-phuong-tay-185250314090514143.htm