Tencent reveals its AI trump card

Author | Huang Yu

Editor | Zhou Zhiyu

At the AI table, as a number of tech giants place high-profile bets on the "most powerful models," pushing the technological competition to a climax, Tencent, holding the top entry ticket, has remained silent strategically, which is disproportionate to its scale.

This "defensive" stance was completely broken this year, as Tencent chose to reveal its cards.

From the early year's "Yuanbao" quickly connecting to DeepSeek for a trial run, to the recently concluded World Artificial Intelligence Conference (WAIC), it has for the first time fully unveiled an AI landscape centered around massive applications, revealing a more proactive and open Tencent.

The press conference at WAIC was less about showcasing technological achievements and more about announcing a shift in strategic focus. Compared to the industry's general obsession with "good technology," Tencent places greater emphasis on creating "good products." Behind this is Tencent's judgment on the ultimate outcome of AI: as technology itself gradually commodifies, the truly implementable "good products" will be the final moat.

Based on this, Tencent's AI strategy is not about promoting a "trump card model" from the top down, but rather starting from users and scenarios, aiming to fully integrate "usable, easy to use, and useful" AI tools into the digital daily lives of 1.4 billion people.

This clear "product-first" approach has also injected a strong boost into Tencent's value reassessment. Since the beginning of this year, its stock price has risen over 32%, and its total market capitalization has returned to over HKD 5 trillion. And this is just the beginning narrative of AI reshaping the industry.

In this ultimate battle for the future, technology is merely the entry ticket; what truly crowns the victor is the real application and choice of billions of users.

Full Coverage of B-end and C-end

After more than two years of effort, Tencent has built a "1+3+N" AI panoramic system, with Tencent's self-developed Hunyuan model as the core engine, constructing a complete architecture that encompasses platform capabilities and diverse applications.

Among them, Tencent has launched three major platform capabilities covering both B-end and C-end, including two major intelligent agent development platforms—"Tencent Yuanqi" for C-end and "Tencent Cloud Intelligent Agent Development Platform" for B-end, as well as the newly released "Tencent Embodied Intelligence Open Platform Tairos" for the robotics field.

Tencent Yuanqi was launched last year, and at this WAIC, it also brought a new upgrade, further integrating resources from the Tencent ecosystem such as public accounts, Tencent Docs, and WeChat Pay MCP, helping users create more user-friendly public account intelligent agents, enabling one-click authorization for matrix account publishing, and pulling knowledge bases from public accounts.

In May of this year, Tencent Cloud announced a comprehensive upgrade of its large model knowledge engine to the Tencent Cloud Intelligent Agent Development Platform. During WAIC, this platform also welcomed a new upgrade, supporting the easy creation of multiple intelligent agents and collaborative interactions with zero coding, making complex scenarios easy to configure. At the same time, the platform connects to mainstream databases and Tencent Docs, providing more comprehensive data sources for more flexible and accurate dialogue responses Eddie Wu, Vice President of Tencent Cloud, Head of Tencent Cloud Intelligence, and Head of Tencent Youtu Lab, stated: "Tencent places great importance on the intelligent agent development platform sector, and the dual intelligent agent platform can meet the construction needs of various applications. Although these two products target different groups, they share the same underlying platform capability stack, achieving integrated collaboration in development, operation, and distribution, which can help different users create effective assistants capable of handling complex scenarios."

Tencent, which aims to "become a partner for all robot manufacturers," officially launched the embodied intelligence open platform Tairos during WAIC. This is the first domestic embodied intelligence software platform that provides large models, development tools, and data services in a modular way, opening up to the robotics industry through a plug-and-play approach.

In the view of Daniel Zhang, Chief Scientist of Tencent and Director of the Robotics X Lab, Tairos can fill the critical software capabilities for robot bodies and application developers, akin to equipping various robots with a "brain," enabling robots to perceive the world, plan tasks, and make autonomous decisions, evolving from passive executors of commands to proactive intelligent agents that adapt to the real world.

In addition to the three major platform capabilities, Tencent also showcased a rich matrix of AI applications and intelligent agent products distributed across multiple industry scenarios.

Tencent's AI application blueprint is divided into two main parts: one is AI-native applications, including the AI assistant "Tencent Yuanbao," the AI smart workstation "ima," and AI search "QQ Browser"; the other is Tencent ecosystem applications that fully integrate AI capabilities, such as WeChat, QQ, QQ Music, Tencent Meeting, Tencent Docs, Tencent Maps, etc.

In Tencent's AI strategy, Yuanbao holds a prominent position as a pioneer in seizing the next generation of traffic entry points. Since the beginning of this year, Yuanbao has rapidly upgraded its capabilities around search, document understanding, and image processing, while deeply integrating into high-frequency usage scenarios like WeChat and QQ.

Another AI-native application from Tencent, ima, focuses on knowledge base scenarios. By integrating "living knowledge base + large models," it addresses the challenges of knowledge fragmentation and the difficulty of ensuring accuracy in massive information, transforming static knowledge into advanced productivity that drives individuals and organizations. Tencent's disclosed data shows that by the second quarter of this year, the shareable private domain content of ima knowledge accounts increased by as much as 208% month-on-month.

Complementing these zero-to-one AI-native applications, future AI will seamlessly integrate with various scenarios, appearing closer to users and industries, which is precisely Tencent's inherent advantage.

For example, as a super application covering hundreds of millions of users, WeChat and QQ have deeply integrated AI capabilities. Mini programs, search, WeChat Reading, and other WeChat open ecosystems have fully integrated AI, and WeChat users can also add the Yuanbao AI assistant as a dedicated friend. Meanwhile, QQ provides diversified AI capabilities in core social scenarios such as chat conversations, search, audio-visual processing, and community management.

Additionally, the newly upgraded QQ Browser, now an AI browser, is equipped with Tencent's Hunyuan and DeepSeek dual models, offering various functions such as AI search, AI browsing, AI office work, AI learning, and AI writing At WAIC, Tencent also showcased over 10 vertical field Agents, covering high-frequency demands in scenarios such as enterprise services, daily life, and office work. Examples include the "Enterprise Marketing Cloud Agent" that automates the entire marketing process, the "Tencent Maps Travel Planning Agent" that generates personalized travel guides with one click, and the "CodeBuddy Agent" that helps developers achieve automatic code completion and diagnostics.

As pointed out in the report "Symbiotic Partners: Top Ten Trends in Artificial Intelligence by 2025," jointly released by Tencent Research Institute and others, AI is undergoing a critical leap from "reasoning" to "action," evolving from a "digital external brain" that assists humans to a "symbiotic partner" that collaborates with humans.

Therefore, in the era of large models, there is a need for leading underlying technology, a one-stop tool platform, and in-depth AI applications and industrial practices. Based on this, Tencent's AI application landscape is fully opening up from the B-end to the C-end, and its AI strategy is expected to become its second growth curve, which is one of the biggest expectations of the capital market for Tencent.

Firm Investment in AI Strategy

This year, Tencent proposed a multi-model strategy of "firmly investing in self-developed models + openly embracing advanced open-source models," which is seen as a pragmatic "two-legged" approach.

The underlying large model capability is key to whether AI applications can truly be effective, serving as a "brain" of sorts. Since the beginning of this year, the iteration speed of Tencent's self-developed Hunyuan model has significantly accelerated. The Tencent Hunyuan self-developed reasoning model Thinker (T1), which excels in complex tasks and deep reasoning, has been rapidly iterating since its launch on the Yuanbao App at the beginning of the year. Additionally, Tencent has released a new generation of fast-thinking model Hunyuan Turbo S, which focuses on faster task processing capabilities.

Based on the Turbo S foundation, Tencent has also launched the visual deep reasoning model T1-Vision and the end-to-end voice call model Hunyuan Voice. Furthermore, a series of multimodal models, including Hunyuan Image 2.0, Hunyuan 3D v2.5, and Hunyuan Game Visual Generation, have also been simultaneously "launched."

During this WAIC, Tencent Hunyuan officially released and open-sourced the industry's first 3D world generation model—Hunyuan 3D World Model 1.0. With this model, users can generate a virtual world that is 360° roamable and editable within minutes by simply inputting a sentence or an image, and output standardized 3D assets that are compatible with mainstream engines, significantly shortening the content production cycle.

In addition to the 3D model, Tencent Hunyuan revealed that a series of models, including edge-side hybrid reasoning language models, multimodal understanding models, and game visual models, have open-source plans in the pipeline.

Since the wave of large AI models began, the debate over "open-source vs. closed-source" has been a hot topic in the industry. The essence of open-sourcing large models is to exchange openness for ecology, to nurture technology through ecology, and to accelerate technological iteration with the power of global developers. Tencent is undoubtedly a firm supporter of open-source. Currently, Tencent's Hunyuan large model has achieved open-source across all modalities, including images, videos, 3D, and text, with the number of image and video derivative models reaching 1,400 and 1,600 respectively, and the community download count for the Hunyuan 3D series models exceeding 2.3 million At the same time, Tencent does not shy away from embracing advanced open-source models. Insiders have indicated that Tencent's AI strategic anchor point has always been the user perspective, specifically how to better connect and utilize the most advanced large model capabilities through Tencent's more user-friendly product interfaces and stronger product power. The integration of DeepSeek by Tencent Yuanbao at the beginning of the year is an example.

"Leading-edge large models require a series of key product capabilities as a bridge to transform technical potential into user value. For instance, selecting the most suitable model based on the scenario, reducing model hallucinations through prompt engineering, and ensuring application stability and availability through high concurrency capabilities. These are precisely the areas where giants like Tencent excel," an industry insider analyzed.

To achieve rapid product innovation and deep model research and development, Tencent has integrated AI products and applications such as Tencent Yuanbao, QQ Browser, Sogou Input Method, and ima into CSIG (Cloud and Smart Industry Group) this year, and has made organizational adjustments to TEG (Technical Engineering Group), which is responsible for the development of Tencent's mixed large model.

Tencent's determination to invest in AI strategy is evident. Financial reports show that in the first quarter of this year, Tencent's capital expenditure reached 27.48 billion yuan, a year-on-year increase of 91%; additionally, general and administrative expenses increased by 36% year-on-year to 33.6 billion yuan, one of the main reasons being the increase in R&D expenses related to AI businesses.

Cai Guangzhong, Vice President of Tencent, stated that AI represents a technological innovation of industrial revolution level, which will bring unprecedented value to both individuals and industrial development. However, relative to a complete technological cycle, AI is still in its early developmental stage.

Regardless, in this heated global AI competition, Tencent is accelerating its pace. In the future, as the dividends of AI technology are further released, Tencent is expected to form a "data-scenario-technology" growth flywheel through the AI panorama it has built, which will be difficult for other companies to replicate