Currently, Chinese technology companies are actively paying attention to open-source artificial intelligence models in the practical plane, launching such digital products at a high-intensity pace, while watchers and insiders claim that the implementation of the appropriate approach contributes to boosting AI adoption and scaling some other innovative functional solutions of the advanced category in the Asian country, at the same time assuming that in this case, there is probably a kind of Android moment for the sector of virtual cognitive systems.
The source of what can be called the open-source shift impulse was China’s startup DeepSeek. This startup launched the R1 artificial intelligence model in the current year, which actually challenged the seemingly in the past unconditional and unequivocal dominance of technologies from companies from the United States. Also, the mentioned AI model has become a kind of doubt factor regarding the need for huge spending of Big Tech on the implementation of the development process of large language models and data centers. According to Jeffries analysts’ preliminary calculations, DeepSeek spent $5.6 million to train R1. This amount is significantly lower than the average reading, which is already perceived by representatives of the machine intelligence industry and investors as a kind of objective financial necessity.
The R1 demonstrated a level of performance that shows the ability of this artificial intelligence model to compete with advanced virtual products of a similar category from US-based companies. According to media reports, the success of DeepSeek has become a source of significant acceleration in the adoption of open-source AI models.
Wei Sun, principal analyst of artificial intelligence at Counterpoint Research, stated that the achievements demonstrated by R1 prove that open-source strategies could lead to faster innovation and broad adoption. Also in this context, it was noted that many companies have implemented the mentioned model of artificial intelligence. Moreover, the analyst stated that R1 is actively reshaping the machine intelligence landscape in China. Apart from that, it was overlined that large companies involved in the relevant process such as Baidu, move to their own large language models in a strategic response.
In the current month, Baidu released a new version of its Ernie 4.5 artificial intelligence model and a new Ernie X1 reasoning model. For individual users, access to these virtual products was provided free of charge. The company also intends to release Ernie 4.5 model series open-source in the summer. It is expected that this debut will take place at the end of June.
According to media reports, citing experts, the mentioned actions and intentions of Baidu related to the adoption of open-source artificial intelligence models reflect a larger process that can be described as a kind of shift in China’s AI sector. In this case, the most fundamental change is that companies are moving away from business strategies focused on proprietary licensing.
Lian Jye Su, chief analyst with Omdia technology research and advisory group, underlined that Baidu has always actively supported its proprietary business model and opposed open-source. At the same time, as the expert noted, disruptors like DeepSeek have proved that open-source artificial intelligence models can be as reliable and competitive as proprietary ones.
Open-source generally means a term implying software in which source code is made freely available on the web for possible modification and redistribution. DeepSeek is not a pioneer in the development of open-source artificial intelligence models. In the United States, similar virtual products have already been elaborated by local technology giants. In the relevant context, it is worth mentioning Llama from Meta and Gemma from Google. At the same time, some experts claim that these artificial intelligence models are not really open source. The argument in favor of this statement is that licenses restrict certain uses and modifications. Also, in this case, an important circumstance is that the training data sets are not public.
R1 is distributed under an MIT License. Wei Sun described this license as one of the most permissive and widely adopted. According to the expert, the corresponding case facilitates unrestricted use, modification, and distribution, including for commercial purposes.
In February, DeepSeek held an Open-Source Week. In this case, the company released additional technical details of the R1 development process. Although this model of artificial intelligence is free, at the initial stage, a fee is provided for the Application Programming Interface, which allows consumers to integrate AI models and their capabilities into other apps. At the same time, the cost of using R1 is significantly lower than that typical for the newest digital products from OpenAI and Anthropic. The mentioned artificial intelligence industry players generate revenue by charging individual users and enterprises for access to some of their AI models. The mentioned models belong to the closed-source category. In this case, datasets and algorithms are not open to the public.
In addition to Baidu, other Chinese technology giants such as Alibaba Group and Tencent are increasingly providing their artificial intelligence services to consumers for free and making more and more open-source AI models.
Alibaba Cloud said last month that it uses its open-source machine intelligence models to generate videos. Also in March, Tencent, as reported by the media, released five new open-source artificial intelligence models that can convert text and images into 3D visuals.
Also in China, the tendency of a kind of open-source shift involves smaller players in the local technology space. In the relevant context, ManusAI is a representative example. The mentioned Chinese company operating in the area of artificial intelligence unveiled an AI agent that allegedly outperforms OpenAI’s Deep Research. ManusAI stated its intention to shift towards open source. The company’s co-founder Ji Yichao said in a product demo video that this would not have been possible without the amazing open-source community, which is why the firm is committed to giving back. ManusAI functions as a multi-agent system powered by several distinct models. Later in the current year, the company is going to open source some of these models.
Zhipu AI, one of the leading startups in the Chinese artificial intelligence industry, announced on WeChat this month that 2025 will be the year of open source.
Ray Wang, principal analyst and founder of Constellation Research, argues that the emergence of DeepSeek was a factor that compelled other technology companies from the Asian country to make the mentioned moves. It was noted that with DeepSeek free, other Chinese competitors cannot charge for the same thing. Ray Wang stated that the mentioned companies have to move to an open-source business model to ensure their competitiveness.
Artificial intelligence scholar and entrepreneur Kai-Fu Lee suggests the mentioned dynamic will have an impact on OpenAI. In this context, it was noted that it would be difficult for the specified startup that developed ChatGPT to justify its prices against the background of free and formidable competition. It was also underlined that the biggest revelation from DeepSeek is that open-source won. Kai-Fu Lee’s startup 01.AI has built a large language model platform for enterprises that seek to use DeepSeek.
OpenAI has not yet made any statements about plans or even the likelihood of abandoning its business model. It’s worth noting that ChatGPT’s debut in November 2022 was actually the initial stage of the so-called artificial intelligence boom, which can be described as a fundamental, systemic component of the functional content of the current era in the history of the technological dimension of the world.
OpenAI was founded in 2015 as a non-profit structure. Currently, the startup is on its way to transforming into a for-profit organization.
Wei Sun claims that DeepSeek and OpenAI represent very different ends of the artificial intelligence industry. It was also noted that in the mentioned industry, a division may persist between open-source players that innovate off one another, and closed-source companies forced to maintain expensive cutting-edge AI models.
The open-source tendency has provoked doubts about the need for huge investments raised, for example, by OpenAI. It’s worth mentioning that Microsoft has invested $13 billion in the ChatGPT developer. Also in January, the media published information according to which OpenAI is negotiating a new round of financing. In this case, the startup plans to raise $40 billion. The valuation of OpenAI is expected to grow to $340 billion.
In September, the media published information according to which OpenAI expects losses of about $ 5 billion with projected revenue of $3.7 billion. Startup chief financial officer Sarah Friar said that revenue of $11 billion is definitely in the realm of possibility for the ChatGPT developer in the current year.
Ray Wang noted that Chinese companies have chosen the open-source route because they compete with the more proprietary approach of US firms. It was also highlighted that technology brands from an Asian country are aiming for faster adoption of their artificial intelligence models than the pace of implementation of a similar process by United States-based companies developing closed AI models.
Tim Wang, managing partner of tech-focused hedge fund Monolith Management, said during a conversation with media representatives that machine intelligence models from companies such as DeepSeek were great enablers and multipliers in China, demonstrating how things can be done with more limited resources. It was also noted that open-source artificial intelligence models have pushed down costs, forming opportunities for product innovation, which is something that Chinese firms have historically been good at. Tim Wang calls the development the Android moment. In this context, he refers to the moment when Google’s Android made its operating system source freely available, contributing to innovation and development in the non-Apple app ecosystem.
Tim Wang noted that the previous assessment of the condition of the Chinese artificial intelligence industry provided that it is 12-24 months away from a similar US industry. According to him, it is now assumed that the corresponding gap is from three to six months.
At the same time, some other experts disagree that open-source artificial intelligence should be considered and perceived as a phenomenon deeply tied to the technological competition between the United States and China. At the same time, the facts realized in the space of objective reality do not correspond in the best way to the mentioned point of view. In this context, it is worth mentioning that some companies based in the United States have integrated R1 and benefited from it.
Alibaba Group Chairperson Joe Tsai, during a conversation with media representatives at a conference in Singapore in the current month, said that the so-called DeepSeek moment is not about whether China has better artificial intelligence than the US or vice versa. It was noted that in this case, the main aspect is the power of open source. Joe Tsai stated that open-source artificial intelligence models provide AI capabilities to everyone from small entrepreneurs to large corporations, which will lead to greater development, innovation, and proliferation of digital intelligence apps.
It is worth noting that against the background of the active development of machine intelligence, the issue of cybersecurity has become more relevant. Scammers also have access to digital intelligence tools. Personal awareness of users is important to counteract the appropriate threat. For example, an Internet search query such as how to know if my camera is hacked will allow anyone to get information about signs of unauthorized access to the device.
