Tencent's WeChat, Baidu Search and Ernie AI Agent Embrace DeepSeek

TMTPOST -- Two Chinese internet giants Tencent Holdings Ltd. and Baidu, Inc. are jumping on the bandwagon to integrate artificial intelligence (AI) upstart DeepSeek’s model into their offerings.

Credit：Pixabay

Tencent’s popular social media platform WeChat is launching a small-scale testing to integrate DeepSeek’s reasoning model R1, Tencent confirmed to multiple Chinese news media outlets at weekends. Selected users in the beta test have the access to the full version of DeepSeek's R1 for free if they click the "AI Search" label at the top of the search bar and choose the “Deep Thinking” mode, Tencent said. The test is only on Chinese WeChat app, or Wexin, rather than the WeChat app that targets overseas users.

The search service involves the provision and deployment of DeepSeek’s open-source large AI models, according to WeChat’s page of introduction to features. The response DeepSeek-R1 provided, featuring a chain-of-thought inference process, quoted content both from posts of official WeChat accounts and web page information, and users can forward it to their WeChat friends and groups.

The DeepSeek-backed WeChat AI search feature not only integrates information sources within the Tencent ecosystem such as WeChat public accounts and video accounts, but also supports online search, which can provide more comprehensive, real-time and accurate answers, Tencent said.

Tencent noted the new feature only integrates official accounts and other public information on the internet, and vowed not to pose privacy risks as the feature will not take advantage of users' personal information or related private information. Tencent will continue to optimize the feature that is under the test based on user experience and feedback.

A number of Tencent’s products, including Tencent Cloud's AI code assistant and Yuanbao, an AI assistant powered by Tencent’s flagship model Hunyuan, are exploring integration with DeepSeek to offer users a richer experience and enhanced services, the company said.

Baidu announced at weekends that its search engine Baidu Search and the large language model (LLM) platform Ernie AI Agent will fully connect to DeepSeek. Baidu will bring the deep search capabilities of DeepSeek service and the Ernie model at no cost for search users, and developers on the Ernie AI Agent platform will also be able to make use of the DeepSeek model for free at any time to create and fine-tune their AI agents.

The free DeepSeek application stunned both Silicon Valley and Wall Street and went viral overseas since late January. DeepSeek’s popular chatbot service app powered by its AI models jumped to the No.1 spot in app stores on January 27, dethroning OpenAI’s ChatGPT as the most downloaded free app in U.S. on Apple Inc.'s App Store, and has maintained its lead since. DeepSeek also topped the global Apple App Store download charts the same day and held the No.1 position since, according to Appfigures data that excludes third-party app stores in China. DeepSeek app was downloaded 16 million times in its first 18 days, surpassing the 9 million downloads recorded by OpenAI's ChatGPT app in the same timeframe, data from Sensor Tower shows. DeepSeek's AI assistant is also the leading free download in Google's App Store, a position it has held since January 28.

As DeepSeek's low-cost, high-efficiency model gains traction, many tech heavyweights accelerated the iteration of their proprietary AI models or pivoted to open-source their AI models.

Google launched its flagship AI model, Gemini 2.0 Pro Experimental, in early February, alongside Gemini 2.0 Flash Thinking. Gemini 2.0 Pro can now access Google Search and execute code on behalf of users. It also features a 2-million-token context window, allowing it to process around 1.5 million English words in a single prompt—enough to read all seven Harry Potter books with 400,000 words to spare.

ByteDance's Doubao AI Foundation team on Wednesday unveiled a new UltraMem sparse model architecture, which improves inference speed by 2–6 times compared to traditional Mixture of Experts (MoE) architectures, while reducing costs by up to 83%. This breakthrough has been accepted by ICLR 2025, a top-tier AI conference, offering new insights into improving AI inference efficiency and scalability.

OpenAI will release GPT-5 in both ChatGPT and its API, and free ChatGPT users will get “unlimited chat access at the standard intelligence setting”, though subject to abuse thresholds, the CEO Sam Altman announced in a post on social media platform X on Wednesday.

Baidu on Thursday said it will make its AI chatbot Ernie Bot for free to use starting April 1. Besides, the Chinese leading search engine operator plans to release an advanced search function on the same day. That function, which is said to include upgraded reasoning capability, will also be available to users at no cost.

According to a Goldman Sachs report, Tencent is among the first Chinese cloud service providers (CSPs) to implement the DeepSeek-R1-671B model, leveraging WeChat’s unique content and Tencent Cloud AI’s infrastructure. Goldman Sachs views this as part of Tencent’s broader AI strategy, aimed at developing a leading AI platform in China with both proprietary (Hunyuan) and third-party models like DeepSeek.

While Tencent and Baidu accelerate AI integration, ByteDance has reflected on its slower response to industry trends. ByteDance CEO Liang Rubo acknowledged that the long-chain reasoning capabilities in DeepSeek-R1 were not a novel breakthrough, as OpenAI had introduced similar innovations last September. He admitted that ByteDance had been too slow to act on these developments.

The rise of DeepSeek has triggered strategic shifts among China's leading AI startups, often referred to as the big model "Six Little Tigers" (Zhipu AI, Baichuan Intelligence, 01.AI, Moonshot AI, MiniMax, and StepAI). Some are investing in independent model development, while others are integrating DeepSeek to enhance their offerings.

01.AI has pivoted away from training massive AI models, instead focusing on lightweight, cost-effective models for industry applications. Their overseas AI platform, PopAi, has already adopted DeepSeek. Additionally, 01.AI launched an AI research base in Suzhou, partnering with key industry players to develop AI solutions for manufacturing, finance, healthcare, and government services.

Other firms, like StepAI and MiniMax, have incorporated DeepSeek into their products. StepAI’s "StepAsk" app and MiniMax’s international version have both integrated the DeepSeek-R1 model. MiniMax CEO Yan Junjie has announced plans to open-source their technology in 2025, believing that an open-source approach would have accelerated their growth from the outset.

Meanwhile, Baichuan Intelligence has doubled down on AI-driven healthcare. In January, they introduced the Baichuan-M1-preview model with language, vision, and search reasoning capabilities. Their "AI Pediatrician" recently participated in China’s first dual-doctor consultation alongside human specialists at Beijing Children's Hospital.

Moonshot AI and Zhipu AI continue refining their large models and AI agent capabilities. Moonshot AI's latest Kimi k1.5 model has attracted industry attention, and reports suggest an even more advanced model is forthcoming. Meanwhile, Zhipu AI has integrated its Agentic GLM into Samsung’s Galaxy S25 series, bringing AI-powered real-time voice and video calls to smartphones.

According to a Morgan Stanley report, AI development is shifting towards differentiation, with companies choosing between premium, high-performance AI and more affordable alternatives. Analysts liken the current AI boom to the internet’s rise in 1995, predicting that market winners will be those who can scale AI technology efficiently and drive commercialization.

With Baidu and Tencent integrating DeepSeek into their flagship products, traditional search may soon be disrupted. According to a survey by the Securities Times, 59% of users now rely primarily on AI-powered search tools, while traditional engines like Baidu hold only a 22% market share.

Tencent clarified that its AI Search feature does not access users’ private WeChat data, instead sourcing information from public accounts and high-quality web content. Analysts believe WeChat's integration of DeepSeek will accelerate AI adoption, particularly among younger users who already favor AI-driven search over traditional methods.

Industry experts predict that 2025 will be a breakthrough year for AI applications. IDC China Research Manager Cheng Yin emphasizes that foundational AI models like DeepSeek are ushering in a new development paradigm, accelerating the commercialization of AI across various industries. From personal productivity tools to business automation, AI is set to redefine how users interact with digital information.

As China’s AI ecosystem matures, companies must choose between building proprietary models or leveraging DeepSeek’s advancements. One thing is clear: the AI race is far from over, and the next phase of innovation will determine the industry's future leaders.