China’s open source hug repeats conventional wisdom around artificial intelligence

China is adopting open source AI models in the observers of a trend market and experts say it is promoting the adoption and innovation of AI in the country, and some suggest that it is a “Android Moment” for the sector.

The open source shift It has been headed by the Deepseek startup, whose model r1 launched earlier this year challenged The American technological domain and asked questions about Big Tech’s mass expenditure on large language models and data centers.

While R1 created a splash in the sector due to its performance and lower cost claims, some analysts say that Deepseek’s most significant impact has been to catalyze the adoption of open source AI models.

“Deepseek’s success shows that open source strategies can lead to faster innovation and wide adoption,” said Wei Sun, a leading artificial intelligence analyst in counterpoint research, noting that a large number of companies have implemented the model.

“Now, we see that R1 is actively restructuring China’s panorama, with large companies such as Baidu Moving for the open source its own LLM in a strategic response, “he added.

March 16, Baidu launched the latest version of his AI Model, Ernie 4.5, as well as a new reasoning model, Ernie X1, making them free for individual users. Baidu also plans To make the series of Ernie 4.5 source models since the end of June.

Experts say that Baidu open source plans represent a broader change in China, far from a commercial strategy that focuses on proprietary licenses.

“Baidu has always supported its patented business model and was a vowel against open source, but disruptors such as Deepseek have shown that open source models can be as competitive and reliable as the owners,” Lian Juye his, main analyst with the Research Group and the OMDIA Technology Advisory Group He told CNBC previously.

The AI ​​models that are called themselves had been open source before the appearance of Deepseek, with GoalHowever, Google’s flame and gemma in the US Some experts He argues that these models are not really open source, since their licenses restrict certain uses and modifications, and their training sets are not public.

Deepseek’s R1 is distributed under a ‘MIT license’, which Sun de Counterpoint describes as one of the most permissive and widely adopted open source licenses, facilitating the use, modification and distribution without restrictions, including commercial purposes.

The Depseek team even had a “Open source week“Last month, which saw him launch more technical details about the development of his R1 model.

While the Deepseek model itself is free, starting positions for the application programming interface, which allows the integration of AI models and their capacities in the applications of other companies. However, their API positions are announced to be much cheaper compared to the latest OpenAi and Anthrope offers.

Operai and Anthrope also generate income charging individual users and companies to access some of their models. These models are considered “closed source”, since their data sets, and the algorithms are not open for public access.

In addition to Baidu, other Chinese technological giants such as Alibaba group and Tencent He has been providing his offers more and more for free and are doing more open source models.

For example, Alibaba Cloud said last month that was open AI models for videos generationWhile Tencent liberated Five new open source models Earlier this month, with the ability to convert text and images into 3D images.

Youngest players are also promoting the trend. Manusai, a Chinese A -firm that recently announced a You have an agent That claims to overcome Openai’s deep research, he said he would change towards the open source.

“This would not be possible without the incredible open source community, so we are committed to return,” said co -founder Ji Yichao in a product Demonstration video. “Manusai operates as a system of multiple agents promoted by several different models, so at the end of this year, we will open some of these models,” he added.

Zhipu AI, one of the main new companies in the country, announced this month in Wechat that 2025 would be “the year of open source.”

Ray Wang, the main analyst and founder of Constellation Research, told CNBC that companies have been forced to make these movements after the appearance of Deepseek.

“With Deepseek Free, it is impossible for any other Chinese competitor to load for the same. They have to go to commercial open source models to compete,” Wang said.

The academic and businessman of the Kai-Fu Lee also believes that this dynamic will affect Openai, pointing out in a recent publication on social networks It would be difficult for the company to justify its prices when competition is “free and formidable.”

“Deepseek’s greatest revelation is that he has won the open source,” said Lee, whose Chinese startup 01.AI has built a LLM platform for companies seeking to use Deepseek.

Openai, who began the frenzy AI when he launched his crawgpt bot in November 2022, has not pointed out that he plans to change his patented business model. The company that began as a non -profit organization in 2015 is moving towards a for profit structure.

Sun says that Operai and Deepseek represent very different ends from the AI ​​space. She adds that The sector could continue to see the division between open source players who innovate with each other and closed code companies that have been pressed to maintain high -cost avant -garde models.

The open source trend has presented to question the massive funds raised by companies such as OpenAI. Microsoft has invested $ 13 billion in the company. Is in conversations to raise up to $ 40 billion in a financing round That would raise its valuation of up to $ 340 billion, CNBC confirmed at the end of January.

Yon September, CNBC confirmed The company expects around $ 5 billion in losses, with revenues linked to revenues of $ 3.7 billion. The financial director of Operai Sarah Friar, has also said that the income of $ 11 billion is “Definitely in the field of possibility“For the company this year.

On the other hand, Chinese companies have chosen the open source route, as they compete with the most patented approach to US companies, Wang de Constellation Research said. “They expect faster adoption than closed models in the United States,” he added.

Talking to CNBC “Street Sigs Asia“On Wednesday, Tim Wang, managing partner of Monolith Management of coverage funds centered on technology, said that companies such as Deepseek have been” great facilitators and multipliers in China “, demonstrating how things can be done with more limited resources.

According to Wang, open source models have reduced costs, opening doors for product innovation, something in which Chinese companies have historically been very good.

He calls the “Android Moment” development, referring when Google’s Android made his Source code of the operating system available for freepromoting Innovation and development In the applications ecosystem without Apple.

“We used to think that China had 12 to 24 months ago (United States) in AI and now we believe they are probably three to six months,” Wang said.

However, other experts have minimized the idea that open source AI should be seen through the Chinese lens and American competition. In fact, several US companies have Integrated and benefited of Deepseek’s R1.

“I think that the so -called deep moment is not about if China has a better thing than the United States or vice versa. It really is the power of the open source,” said Alibaba Group president Joe Tsai, to the Converge Conference of CNBC in Singapore earlier this month.

Tsai added that open source models give the power of AI to all, from small entrepreneurs to large corporations, which will lead to greater development, innovation and a proliferation of AI applications.