Chinese open code covers conventional wisdom around artificial intelligence

China covers open code AI models in trends market observers, and internal persons say it enhances the acceptance of AI and innovation in the country, some suggest that it is a “Android moment” for the sector.

Open code replacement is headed by AI startup Deepseek, whose model R1 was released earlier this year contested American technological dominance and raised questions about Big Tech’s mass costs about large language models and data centers.

While the R1 has created spraying in the sector due to its performance and claims for lower costs, some analysts say that the most significant impact of Deepseek is in catalyzing the acceptance of open source AI models.

“Deepseek’s success proves that open source strategies can lead to faster innovation and wide acceptance,” says Wei Sun, Chief Analyzer of Artificial Intelligence at Counterpoint Research, noting that a large number of companies have implemented the model.

“We now see R1 actively reshapes the Chinese landscape of AI, with large companies such as Baidu Moving to Open Source her own LLM in a strategic answer, “she added.

On March 16, Baidu released the latest version of its AI Model, Ernie 4.5, as well as a new model of reasoning, Ernie X1, which makes them free for individual users. Baidu also plans To make the Ernie 4.5 series with an open code of End-June.

Experts say the Baidu open code plans are a broader change in China, far from a business strategy that focuses on its own licensing.

“Baidu has always supported its own business model and has been vowed against open source, but destroyers like Deepseek have proven that open source models can be as competitive and reliable as their own,” Lian G, Chief Analyzer with Technology and Advisory Eligo Omdia He told CNBC earlier.

AI models called open code have existed before the appearance of Deepseek, with Meta“Google’s Ellama and Jemma are the main examples in the US, however, some experts He claims that these models are not really an open code, as their licenses limit certain applications and modifications and their training sets are not public.

Deepseek’s R1 is distributed under the “MIT license”, which Counterpoint’s Sun describes as one of the most daring and widely accepted open source licenses that facilitate unlimited use, modification and distribution, including commercial purposes.

Deepseek’s team even heldOpen Code Week“Last month, in which he was seen more technical details about the development of his R1 model.

While the Deepseek model itself is free, the startup fees for the application programming interface that allows the integration of AI models and their capabilities into applications of other companies. However, his API fees are advertised as far more expensive than the most offers of Openai and Anthropic.

Openai and Anthropic also generate revenue by charge individual users and businesses to access some of their models. These models are considered a “closed code” as their data sets and algorithms are not open to public access.

In addition to Baidu, other Chinese technology giants such as Alibaba Group and Tencent They are increasingly providing their AI offers free of charge and make more open source models.

For example, Alibaba Cloud said last month that it was open to a source AI models to generate videoWhile the messages have been released Tencent Five new open source models Earlier this month with the possibility of transforming text and images into 3D visualizations.

The younger players also continue the trend. Manusai, a Chinese AI company that recently revealed You have an agent This claims that he is superior to deep Openai research, he said he would target open code.

“This would not have been possible without the incredible open code community, which is why we are committed to returning co -founder GGYCA to a product Demo videoS “Manusai works as a multi-agent system powered by several different models, so we will open some of these models later this year,” he added.

Zhipu AI, one of the leading AI startups in the country, announced this month to WeChat that 2025 will be the “Open Code Year”.

Ray Wang, chief analyst and founder of Constellation Research, told CNBC that companies were forced to make these moves after the appearance of Deepsek.

“With Deepseek Free, it is impossible for all other Chinese competitors to charge the same thing. They have to move to open source business models to compete,” Wang said.

AI Scholar and entrepreneur Kai-Fu Lee also thinks Social media post that it would be difficult for the company to justify its pricing when the competition is “free and great”.

“The biggest disclosure from Deepseek is that it has won an open source,” says Lee, whose Chinese startup 01.Ai has built a LLM platform for businesses that want to use Deep.

Openai – who started Ai Frenzy when he released his Chatgpt Bot in November 2022 – does not signal that he plans to shift from his own business model. The company that launched as a non -profit purpose in 2015 moves to profit Structure.

Sun says Openai and Deepseek represent many different ends of AI spaceS She adds that The sector can continue to see a division between open source players that innovate each other and closed -code companies that are under pressure to maintain high avant -garde models.

The open code trend has asked the huge funds raised by companies like Openai. Microsoft has invested $ 13 billion in the companyS In conversations is to raise up to $ 40 billion in financing This would raise its estimate to $ 340 billion, CNBC confirmed in late January.

IN September, confirmed CNBC The company expects losses of about $ 5 billion, with revenue attached to revenue of $ 3.7 billion. The Otraki Financial Officer Sarah Sarah also said revenue of $ 11 billion are “definitely in the sphere of possibility“For the company this year.

On the other hand, Chinese companies have chosen the open source route as they compete with the more owned approach of US companies, said Wang on Constellation Research. “They are hoping for a faster reception than closed US models,” he added.

Speaking to CNBC “Street signs Asia“On Wednesday, Tim Wang, a managing partner of the Monolith Management technologically focused hedge fund, said models from companies like Deepseek are” big factors and multipliers in China, “demonstrating how things can be done with more limited resources.

According to Wang, open source models have reduced the costs, opening doors for innovation of products-something, which he says, that Chinese companies were historically very good.

He calls the “Android Moment” Development, citing when Google Android made his The operating system source code freely availablepromotion Innovation and Development In the ecosystem of the app, which is not Apple.

“We thought China was behind 12 to 24 months (USA) in AI and now we think it was probably three to six months,” Wang said.

However, other experts have downplayed the idea that open code AI should be seen through the China lens and the US competition. In fact, several US companies have integrated and favored From R1 to Deepseek.

“I think the so-called Deepseek moment is not about whether China has better II than the United States or back. It’s actually about the power of the open code,” Alibaba Joe Tsai group chairman told CNBC in Singapore earlier this month.

Cai added that open source models give the power of AI to all, from small entrepreneurs to large corporations, which will lead to more development, innovation and distribution of AI applications.