【By Chen Sijia, Observer Net】"China has surpassed the United States in the global open-source artificial intelligence (AI) model market." According to a report by the Financial Times on November 26, a study by the Massachusetts Institute of Technology (MIT) and the open-source AI startup Hugging Face found that in the past year, the share of downloads of open-source AI models developed by Chinese teams has risen to 17%, surpassing their American counterparts for the first time, gaining a key advantage in the global application of AI technology.

Amid the AI boom sweeping the global tech industry, American tech giants such as OpenAI, Google, and Anthropic tend to adopt a "closed" strategy, maintaining full control over advanced AI technologies and profiting through user subscriptions and corporate collaborations. In contrast, Chinese tech companies tend to favor more open strategies, releasing a series of open-source models.

Open-source AI models refer to AI models whose code and architecture are publicly available, allowing any developer to download, use, modify, and distribute them. The widespread use of open-source models will have a significant impact on the future development of AI, making it easier for startups to develop products and researchers to improve them.

Through the relentless efforts of Chinese researchers, Chinese open-source models have demonstrated strong competitiveness. A study by MIT and Hugging Face found that in the past year, the share of newly developed open-source AI models from China in the total number of open-source model downloads has risen to 17%, exceeding the 15.8% share of models developed by American teams, marking the first time that Chinese teams have surpassed their American counterparts in this metric.

Wendy Chang, a senior analyst at the Mercator Institute for China Studies in Germany, pointed out: "Compared to the US, open-source is more of a mainstream trend in China. US companies are not willing to do this; they rely on these high valuations to make money and do not want to disclose their commercial secrets."

According to data from MIT and Hugging Face, DeepSeek and Qwen from Alibaba Cloud are the most downloaded Chinese open-source models. The Financial Times reported that the DeepSeek-R1 model shocked Silicon Valley. This model uses low costs and computing power but its performance is sufficient to rival top US models, raising doubts about whether US AI labs can still maintain their competitive edge.

Download share of open-source models according to Hugging Face - Screenshot from the Financial Times report

Shayne Longpre, a researcher at MIT, said that Chinese tech companies are changing the paradigm of AI model releases, with many Chinese companies releasing models every week or every two weeks, offering various versions for users to choose from. In contrast, US tech companies often release a series of models every six months or a year.

Industry insiders told the Financial Times that despite a series of measures such as chip export controls imposed by the US, China has a large number of excellent talents who have shown strong creativity in developing open-source models.

To win in the so-called "AI race," the Trump administration is trying to persuade US companies to invest in open-source models that embody "American values." However, the number of large independent development teams for open-source models in the US is far fewer than in China, and even US tech giants like Meta now choose to increase investment in closed-source models.

In August this year, OpenAI released its first batch of "open weights" models. These models are free to access, but the information provided is less comprehensive than that of open-source models, and they do not include the code and training data needed to train models from scratch.

Janet Egan, a senior researcher at the Center for a New American Security, said: "China's major progress in the field of open-source AI models should be of concern to the US."

The Economist magazine published an article in August stating that although Chinese open-source models are low-key, they have gained increasing popularity among start-up tech companies. Martin Casado, a partner at the US venture capital firm Andreessen Horowitz, said: "I estimate that 80% of current startups are using Chinese open-source models."

The Economist analyzed that from a purely business perspective, proprietary models from US companies generate significantly higher revenue than open-source models from Chinese companies, and it is easier to profit from proprietary models, with the income being reinvested into innovation.

However, open-source models are not only about business achievements. Percy Liang, co-founder of the tech company Together AI, stated that enterprises, governments, and researchers can more easily adjust open-source models, applying them to various specific scenarios, and helping users run AI tools locally without relying on the cloud. Through services such as customized support, open-source models can also generate revenue.

"While US AI laboratories bet on gaining huge profits by pushing smart breakthroughs, Chinese open-source model competitors focus more on promoting the widespread application of AI," concluded The Economist. "If they succeed, the impact of DeepSeek may just be the beginning."

This article is an exclusive piece from Observer Net. Reproduction without permission is prohibited.

Original: https://www.toutiao.com/article/7576968777120711206/

Statement: The article represents the views of its author. Please express your opinion by clicking the [Up/Down] buttons below.