The Chinese artificial intelligence company DeepSeek has shocked markets by claiming that, while utilizing fewer sophisticated computer chips and consuming less energy, its most recent AI model, R1, performs on par with OpenAI’s.
The rise of DeepSeek has sparked worries that, despite limitations on its access to the most cutting-edge technology, China might have surpassed the United States in the battle for artificial intelligence. It is only one of several Chinese businesses developing AI to overtake the United States in the race for technical dominance and make China the global leader in the sector by 2030.
Like the United States, China is spending billions on AI. This week, it established an AI investment fund worth 60 billion yuan ($8.2 billion), only days after the United States imposed new chip export curbs.
To get around restrictions on its access to industry leaders’ cutting-edge computer chips, Beijing has also made significant investments in the semiconductor sector. In addition to plans to create AI academies and include AI education in the curricula of elementary and secondary schools, businesses are providing talent programs and subsidies.
China has put in place laws about AI that cover ethics, privacy, and safety. The topics that the AI models can address are likewise governed by the ruling Communist Party; DeepSeek modifies its responses to comply with these restrictions.
An outline of some further top AI models in China may be found here.
Cloud Qwen-2.5-1M by Alibaba
Alibaba Cloud’s Qwen-2.5-1M is an open-source AI series from a major e-commerce company. Its language models can converse in greater depth and handle very long inquiries with ease. The system is improving its understanding of difficult concepts such as logic, conversations, and code comprehension.
Similar to its competitors, Alibaba Cloud has made its Qwen chatbot—known in China as Tongyi Qianwen—available to the general public. Alibaba Cloud’s AI model suite, including the Qwen2.5 series, has primarily been used by developers and business clients, including banks, video game developers, retailers, and automakers, to develop new products and customize user experiences.
The Ernie Bot 4.0 on Baidu
The first AI chatbot to be made publicly available in China was Ernie Bot, created by Baidu, the country’s top search engine. According to Baidu, the model was made publicly available so that it could gather a lot of actual human input and increase its capabilities.
As of June 2024, there were over 300 million users of Ernie Bot 4.0. Ernie Bot allows users to ask it questions and have it produce images in response to text instructions, much like OpenAI’s ChatGPT.
The Doubao 1.5 Pro from ByteDance
ByteDance, the parent company of TikTok, introduced the AI model Doubao 1.5 Pro last week. With 60 million monthly active users, Doubao is currently among the most well-known AI chatbots in China.
According to ByteDance, the Doubao 1.5 Pro outperforms ChatGPT-4o in the areas of reasoning, coding, knowledge retention, and Chinese language processing. ByteDance claims that because Doubao has a highly optimized architecture that strikes a compromise between performance and lower computational demands, the model is also more affordable and requires less hardware than other large language models.
The Kimi k1.5 from Moonshot AI
Following its most recent fundraising round, Beijing-based startup Moonshot AI was valued at over $3 billion. It claims that its recently launched Kimi k1.5 performs on par with or better than the OpenAI o1 model, which is intended to think more deeply before reacting and is capable of handling more challenging and complex tasks. According to Moonshot, Kimi performs better than OpenAI o1 in coding, maths, and the comprehension of text and visual inputs like images and videos.
