China's AI Landscape Shift.

Author: Lin Zhijia, Titan Media AGI

Image Source: Generated by Boundaryless AI

Following Tencent's internal testing of "AI Search" in WeChat and the integration of the DeepSeek model, Baidu quickly responded.

Titan Media AGI learned that on the evening of February 16, Baidu Search announced that it would fully integrate the latest deep search capabilities of the DeepSeek and Wenxin large models. At the same time, to serve a wide range of developers in creating and fine-tuning intelligent agents using various model capabilities, the Wenxin intelligent agent platform will fully integrate DeepSeek.

Meanwhile, Tencent confirmed on the 16th that WeChat has launched the "AI Search" feature and is officially conducting a gray-scale test of the DeepSeek-R1 model, providing "deep thinking" services.

Tencent stated that some users who qualified for testing can see the "AI Search" label at the top of the WeChat chat window. By clicking on it, they can use the full version of the DeepSeek-R1 model for free, enjoying a more diversified search experience. If the entry is not displayed, it indicates that this gray-scale test has not yet covered that user account, and the WeChat team is gradually expanding the testing scope, asking everyone to patiently wait for further openings. Additionally, several Tencent products, including Tencent Yuanbao, QQ Browser, and QQ Music, have integrated the DeepSeek model.

However, less than a day after its launch, WeChat began to experience "full load." Around 11 PM on February 16, Titan Media AGI learned that users using WeChat's "AI Search" began to see the message "Sorry, the service is busy, please try again later."

Goldman Sachs' latest research report pointed out that Tencent is one of the first Chinese CSPs (Cloud Service Providers) to implement the deep thinking mode of the DeepSeek-R1-671B model and achieve search functionality. By leveraging Tencent's unique WeChat content to enhance reasoning capabilities, supported by Tencent Cloud AI's reasoning infrastructure, Goldman Sachs believes this highlights Tencent's multiplicative strategy in the AI open platform, aiming to create "killer" applications and an AI Agent ecosystem in China.

At the same time, the DeepSeek craze has left other companies feeling a "low tide."

ByteDance CEO Liang Rubo recently reflected on DeepSeek, believing that one of the innovations of the DeepSeek R1 long-chain thinking model is not a first in the industry. After OpenAI released the long-chain thinking model in September last year, becoming a hot topic, ByteDance realized the significant technological changes but felt they did not keep pace. If they had paid attention at that time, they might have had the opportunity to implement it earlier; meanwhile, the "Six Little Tigers of Large Models" (Zhizhu AI, Baichuan Intelligence, Jueyue Xingchen, Lingyi Wanyi, Moon's Dark Side, MiniMax) have gradually differentiated and made different development choices.

Two years after the ChatGPT craze swept through, China's AI industry has entered a new "shift."

The Second Half of Large Models: "Six Little Tigers" Accelerate Differentiation

Entering 2025, a new wave of DeepSeek is coming. Recent Bloomberg data surveyed seven AI industry figures, showing that the estimated valuation of DeepSeek ranges from $1 billion to over $150 billion, with a median valuation range of $2 billion to $30 billion.

According to the Bloomberg Billionaires Index, if valued as above, Liang Wenfeng, who holds 84% of the shares, would have a net worth of $126 billion, placing him among Asia's wealthiest tech tycoons, potentially surpassing Nvidia CEO Jensen Huang, who has a net worth of $118 billion.

According to Titan Media AGI statistics, Anthropic, founded by former OpenAI employees and backed by Google and Amazon, has a valuation of $60 billion; Mistral AI, founded by researchers from Google and Meta, has a valuation of $6 billion. Meanwhile, domestic counterpart Zhizhu AI completed a new round of financing of 3 billion yuan last year, with a pre-financing valuation of 20 billion yuan. However, DeepSeek is far ahead, with a company valuation higher than the total of the "Six Little Tigers of Large Models."

Now, under the impact of DeepSeek, the "Six Little Tigers of Large Models," which have a valuation of up to 20 billion yuan, stand at a new crossroads, making different development choices: some choose to continue investing in the research and development of new models, exploring more possibilities for industrial implementation; others choose to embrace the DeepSeek model, leveraging its advantages to expand new business territories.

First is Lingyi Wanyi.

Before the launch of DeepSeek-R1, Lingyi Wanyi CEO and Innovation Works Chairman Kaifu Lee publicly stated that the company would no longer pursue training super large models, as a model with moderate parameters, excellent performance, faster reasoning speed, and lower reasoning costs is more suitable for commercial scenarios. "It will become a catalyst for the explosion of AI-First applications," and Lingyi Wanyi's overseas AI application PopAi has integrated the DeepSeek model.

On February 14, the industrial large model base jointly established by Lingyi Wanyi and Suzhou High-tech Zone was officially inaugurated. Lingyi Wanyi revealed that the base will focus on creating industry large model solutions in multiple fields such as manufacturing, finance, healthcare, and government affairs, collaborating with leading industry chain partners such as Zhongxi Soft Group, Super Media Group (formerly Modern Communication Group), Innovation Qizhi, Beiyang Quantitative, Chengyuan Technology, Qiongche Technology, and Sui Rui Technology to explore the industrialization path of large model technology from the laboratory to the production line.

Kaifu Lee stated that at this critical juncture of AI technology reconstructing industries, large models are not "castles in the air," but the core engine driving the real economy. With the continuous improvement of foundational model performance, applications will flourish, providing unprecedented opportunities for Chinese teams. The year 2025 is the year of the explosion of AI-First applications, and Suzhou, with its solid industrial foundation and rich application scenarios, is the best testing ground for the implementation of large industrial models.

Currently, Lingyi Wanyi has begun exploring the industrialization of large model capabilities in retail, finance, gaming, and energy sectors, engaging in deep cooperation with leading companies, including Fortune Global 500 enterprises. Their large model ToB solutions have also gained recognition from clients such as China Mobile, Alibaba Cloud, Huawei, Yum China, SF Technology, Kidswant, Meitu, and Feishu.

Earlier this year, Lingyi Wanyi announced the establishment of a "Joint Laboratory for Industrial Large Models" in collaboration with Alibaba Cloud: the Ali Tongyi series of large models will serve as a "teacher model" with strong general capabilities; Lingyi Wanyi will possess internationally advanced, cost-effective model capabilities, enabling agile batch training of vertically directed industrial large models, jointly accelerating the industrial implementation of large models and broadening the ecological prospects of large models.

Alibaba Cloud CTO Zhou Jingren stated that the deep integration of large models with industries is an essential path for China to enter a fully intelligent era. The collaboration between Alibaba Cloud and Lingyi Wanyi to establish the "Joint Laboratory for Industrial Large Models" aims to accelerate the empowerment of the real industry through the co-evolution of large and small models, enriching the large model application ecosystem across various industries. Suzhou, with its strong industrial foundation and rich implementation scenarios in manufacturing, finance, and healthcare, is an excellent base for incubating innovative applications of industrial large models.

Secondly, both Jueyue Xingchen and MiniMax have begun integrating the DeepSeek model.

On February 16, Jueyue Xingchen's latest "Yuewen" app integrated the DeepSeek-R1 model; at the same time, MiniMax 01's overseas version has launched the DeepSeek-R1 deep thinking mode.

MiniMax founder and CEO Yan Junjie stated that MiniMax's plan for 2025 is "open source." "If I could choose again, I should have gone open source from day one. Because open source can accelerate technological evolution."

Meanwhile, Baichuan Intelligence has recently continued to "invest" in the AI healthcare sector. On January 25, they released the new model Baichuan-M1-preview, which possesses language, vision, and search reasoning capabilities. On February 13, the "AI Pediatrician" built on the Baichuan-M1 model went live in Beijing after nearly a month of internal testing.

Reports indicate that on that day, Beijing Children's Hospital conducted the country's first "AI Pediatrician + Multidisciplinary Expert" dual medical parallel multidisciplinary consultation. Participants included 13 experts from various departments, as well as the "AI Pediatrician" jointly developed by the hospital, Baichuan Intelligence, and Xiaoyuan Health Technology (a medical data company invested by Baichuan).

The attendees conducted a multidisciplinary consultation for a child with a skull base tumor accompanied by twitching symptoms, while an engineer input the patient's main complaints and medical history into the model. "The AI Pediatrician also provided suggestions that highly matched the expert group's consultation results."

Finally, Moon's Dark Side and Zhizhu AI continue to focus on model and Agent applications.

On January 25, just hours after the release of the DeepSeek-R1 model, Moon's Dark Side launched the Kimi k1.5 multimodal thinking model, attracting attention. Additionally, there are reports that Moon's Dark Side will soon release a large model that "continues to achieve SOTA results."

In early February, OpenAI's latest paper "Competitive Programming with Large Reasoning Models" stated that two Chinese AI companies independently discovered the secret of o1, referencing both the DeepSeek-R1 and Kimi k1.5 models in the introduction, claiming that these two Chinese AI companies independently discovered the secret of o1. Furthermore, OpenAI pointed out that DeepSeek-R1 and Kimi k1.5 enhance the performance of large models in mathematics and programming through CoT.

As a counterpart to OpenAI, Zhizhu AI will not integrate the DeepSeek model for now but will focus on enhancing Agent applications. Their latest release, the Agentic GLM (a system-level large model developed specifically for mobile), has landed on Samsung's latest Galaxy S25 series phones, providing AI-based real-time voice and video calls, as well as visual understanding, system function calls, AI search, and copywriting capabilities. Additionally, Zhizhu is collaborating with the AI drawing application "Nia Ta."

International investment bank Morgan Stanley's latest report predicts that the AI market is heading towards differentiation. DeepSeek has changed the narrative of the AI industry, demonstrating the rewards of unconventional paths, and the notion that only a few companies can meet the conditions to provide the momentum for AI development with extremely powerful chips and infrastructure is no longer true; companies capable of paying the entry costs are no longer limited to a few industry leaders.

Morgan Stanley pointed out that the DeepSeek craze has prompted some companies to be willing to pay high prices to maintain their leading positions; however, for another camp, AI is a cost center, and for these companies, the pursuit is for more affordable tokens.

"We believe that the market winners will ultimately be the companies that can rapidly scale and commercialize new technological breakthroughs. As the market enters a phase of push, we are gradually optimistic about AI large model participants and believe that starting in the second half of 2025, there will be better opportunities arising from the recovery of edge devices." Morgan Stanley believes that the current AI craze is somewhat akin to the internet boom of 1995, when everyone expected Cisco, AltaVista, or Hotmail to win the race, but Amazon ended up being the ultimate winner by offering the lowest cost per second service tokens.

Traditional Search Will Be Transformed by AI

With Baidu's core and flagship product—Baidu Search—and Tencent's core WeChat both integrating the DeepSeek model, this may signify a reshuffling of the traditional search market, and people's search methods and experiences will soon undergo a complete transformation.

According to the Securities Times, Tencent has provided further clarification on some related details. The explanations are as follows:

• 1. Does AI search include data from public accounts? The DeepSeek integrated into WeChat AI search supports online search (users do not need to select manually), and based on the rich content of public accounts and other high-quality content from the internet, it can provide users with more comprehensive and high-quality answers.

• 2. Is AI search fully operational? Currently, this capability is still in gray-scale testing and will continue to be optimized based on user experience and feedback.

• 3. Why does WeChat's search scenario need to integrate large models? Large models can enhance the intelligence and accuracy of searches, such as better understanding users' search intentions and analyzing and processing complex query content. In line with user needs, Tencent has integrated large models, including Hunyuan and DeepSeek, into the search scenario to further enrich users' search experiences.

• 4. Will AI search use my personal information from WeChat, such as Moments and chats? AI search only integrates public account information and other publicly available information from the internet and will not use users' personal information or related privacy information.

According to the Shanghai Securities Journal, renowned economist and member of the Ministry of Industry and Information Technology's Information and Communication Economic Expert Committee, Pan Helin, stated that for WeChat, integrating DeepSeek is not difficult, especially since Tencent has sufficient computing power to support a larger volume of access. "Recently, DeepSeek has frequently encountered server congestion issues due to high access volume, and WeChat's move provides consumers with an alternative solution."

Public information shows that, according to a survey, 59% of netizens primarily use AI search tools, while traditional searches like Baidu account for only 22%. The integration of WeChat may accelerate this trend, especially appealing to younger users.

Pan Helin believes that WeChat's integration is also beneficial for DeepSeek, as WeChat expands DeepSeek's user reach and reduces its computing power burden.

Zhang Lu, founding partner of Fusion Fund and a Silicon Valley investor, recently stated that today's college students, especially freshmen and sophomores, and even some high school students, spend a significant amount of time using AI tools daily. They spend about 70% to 80% of their time using AI applications on their phones. For instance, many students have almost stopped using traditional Google search and have turned to platforms like ChatGPT and You.com for their searches. This indicates that the relationship between humans and AI may be changing faster than we imagine. From initial misunderstanding and resistance to gradual collaboration, and now to reliance, in the future, AI may become a part of daily life like mobile phones, and new habits will form.

IDC China Research Manager Cheng Yin stated that for AI applications, the updates and upgrades of large models will help accelerate the innovation and commercialization of application scenarios. In the future, whether it is applications aimed at enhancing personal productivity, such as copywriting and content generation, online meeting summaries, AI assistants, and search, or scenarios related to customer service and marketing, or the commercialization of industry-specific scenarios, these will be key areas of market focus this year.

Cheng Yin emphasized that DeepSeek leads foundational large models to open a new paradigm of development. In 2025, the industry will pay more attention to the implementation of large models and generative AI, and the entire ecosystem should work together to accelerate the innovation and commercialization of application scenarios.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。