Tencent Shifts to Aggressive AI Strategy, Yao Shunyu Takes Helm of Foundation Model Division to Seize Leadership in the 'Agent' Era

Tencent (00700-HK) has recently shown unprecedented aggressiveness in the AI domain. With deep internal organizational restructuring and the intensive launch of multiple 'Agent' products, Tencent is accelerating resource integration, aiming to shift AI from technical research to large-scale productivity deployment.

Yao Shunyu Secures Leadership of Hunyuan

To enhance collaborative efficiency in model development, Tencent officially ended its dual-track system on July 23, merging its multimodal model department and large language model department to establish the 'Foundation Model Department'.

This department is unified under Tencent's Chief AI Scientist Yao Shunyu, whose authority has expanded from large language models to overseeing foundational model R&D across language, multimodal, reinforcement learning, and Agent technologies. This move signifies that the Hunyuan model has entered a phase of consolidated resource integration, aiming to build a comprehensive model matrix covering image, video, audio, and 3D generation.

Full-Scenario Layout: From Miora to WorkBuddy

Tencent's commitment to Agents is rapidly materializing in its product lineup. On July 22, Tencent launched Miora, China's first multi-Agent collaborative creative studio. Users need only submit a request, and the system automatically orchestrates different Agents to complete image, video, and even 3D creation, significantly lowering the barrier to professional content creation.

In the office domain, WorkBuddy has performed exceptionally well, surpassing 20 million monthly visits in June, firmly securing the top position in China's PC-based AI-native office Agent market. WorkBuddy has already launched as a standalone app, is available on the HarmonyOS, and has even partnered with Li Weike to release the X-AI Memory Glasses, extending the AI interface into wearable devices and addressing the traditional AI shortcoming of lacking contextual memory.

Breaking Through Bottlenecks in Cost-Effectiveness and Enterprise Deployment

Unlike the industry trend of stacking parameters, the Hunyuan Hy3 model follows a platform-oriented path emphasizing 'cost-effectiveness'. Wu Yunsheng, Vice President of Tencent Cloud, pointed out that Chinese vendors aim for higher cost-efficiency while ensuring intelligence, as enterprises prioritize inference costs over marginal performance gains during commercialization.

However, the biggest obstacle to Agent deployment is no longer technical but the challenge of integrating with enterprises' legacy systems. Wu emphasized that many outdated systems lack standardized interfaces, making it difficult for Agents to be directly invoked. To address this, Tencent launched the ADP enterprise-grade Agent development platform, offering full-process observability and auditability to meet the strict security and compliance requirements of industries such as finance and government.

Building the App Store of the Agent Era

Tencent's ultimate goal is to build a thriving skills ecosystem. Its SkillHub currently hosts 78,000 AI skills, paired with the SkillPay payment system, forming a closed-loop business model where developers publish skills and users invoke and pay for them.

Tencent stated that the B2B market cannot be conquered alone and will continue to drive large-scale replication of Agents by providing underlying platforms and model capabilities in collaboration with industry partners.