Korean Media: China Surpasses the US in Open-Source Artificial Intelligence!

On December 22, South Korean media outlet The Pioneer Economy published an article stating that, led by companies such as DeepSeek and Alibaba, China has surpassed the United States in the global artificial intelligence (AI) open-source market.

A study jointly released by the Massachusetts Institute of Technology and the AI open-source platform HuggingFace showed that in the past year, the share of downloads for newly released AI open-source models from China increased to 17%. This figure exceeded the combined download share of American developers including Google, Meta, and OpenAI, which was 15.8%.

This is the first time that Chinese companies have surpassed American companies in the global AI open-source model market. China's surpassing of the United States in the global AI open-source model market is seen as a significant advantage for the widespread application of Chinese AI open-source models worldwide. Open-source models allow developers to freely download and modify code, lowering the barriers to entry for services and R&D.

This shift comes at a time when the Trump administration is striving to outperform China in the AI race, with the Trump administration encouraging American companies to invest in open-source models that reflect "American values."

The Financial Times analyzed that the open-source strategy of Chinese companies contrasts sharply with the closed-loop model adopted by most major American tech companies.

Large American tech companies such as OpenAI, Google, and Anthropic tend to focus on developing cutting-edge models, maintaining full control over cutting-edge technology, and generating revenue through customer subscriptions or enterprise transactions. Meta has been actively opening up its AI models, such as "Rama," but has recently expanded the use of closed-loop models to develop "super intelligence."

In contrast, Chinese companies such as DeepSeek and Alibaba are more widely opening up their models. The export restrictions imposed by NVIDIA on cutting-edge AI semiconductors have effectively accelerated collaboration within China and stimulated the development of the open-source ecosystem.

MERICS, a European think tank, pointed out: "Open-source is more mainstream in China than in the United States. American companies obtain income through high valuations, so they are reluctant to open-source their technologies."

China's open-source model downloads mainly come from models such as DeepSeek R1 and Alibaba's Qwen. OpenAI released its first 'open-weight' model in August last year, but due to limited training code and training data provided, its applications were restricted.

An expert from MIT explained: "Chinese companies such as DeepSeek and Alibaba have already achieved a paradigm shift in model releases. Unlike American companies that release a series of models every six months or a year, Chinese companies release models every week or every two weeks and provide various variants for users to choose from."

U.S. export restrictions on cutting-edge AI semiconductors have enabled Chinese AI researchers to adopt more creative model development methods, such as using distillation techniques to create smaller yet more powerful models. They are also actively developing AI video generation models.

Experts believe that the number of companies in the U.S. engaged in open-source development is far fewer than in China.

Jeanette Egan, Deputy Director of the Center for National Security and Foreign Policy at CNAS, stated: "The U.S. should pay attention to China's significant progress in the open-source model."

Original: toutiao.com/article/1852208574524496/

Statement: This article represents the views of the author."