Tencent Unveils Hy3 Model and Eyes DeepSeek Investment

Tencent Holdings Ltd. unveiled a preview of Hy3, an upgraded foundational AI model focused on complex reasoning and coding, on April 23, 2026. The model is integrated into the company's chatbot, coding tools, and the QQ instant-messaging platform. This release follows the recruitment of former OpenAI researcher Yao Shunyu as chief scientist for AI and a restructuring of the company's research team to improve training data.

To maintain competitiveness against rivals such as Alibaba Group and ByteDance, Tencent pledged to increase its AI investments to more than $5 billion in 2026. The company is shifting its focus toward agentic AI and the development of task-automation agents within its flagship application.

Simultaneously, Tencent and Alibaba Group entered advanced discussions to invest in the AI startup DeepSeek, which is owned by High-Flyer Capital Management. DeepSeek is reportedly seeking funding at a valuation exceeding $20 billion. The two tech giants intend to integrate DeepSeek's models into their cloud infrastructures to offer low-cost AI services to enterprise clients and create a national champion capable of competing with U.S. firms despite semiconductor restrictions. On April 24, DeepSeek further expanded its offerings by launching V4 Flash and V4 Pro Series LLMs, which utilize a Hybrid Attention Architecture to improve reasoning and long-conversation memory.