On April 23, 2026, Tencent officially launched and open-sourced the Hunyuan Hy3 preview language model. This model is a hybrid expert model that combines fast and slow thinking. It boasts a total of 295 billion parameters, with 21 billion active parameters, and supports a context length of 256K. As the inaugural model trained following Hunyuan's restructuring, it showcases notable enhancements in complex reasoning, instruction adherence, contextual learning, coding, agent capabilities, and inference efficiency. Leveraging the widely-adopted open-source framework vLLM, BIREN Technology successfully executed Day0 adaptation and inference validation for the Hy3 preview model. The synergy between the Hy3 preview model and BIREN's BR166 series GPUs—a collaboration of "domestic model + domestic chip"—will maximize the efficiency of large-scale model deployment. This partnership provides developers and industry users with a more efficient and stable deployment experience, facilitating a wide range of real-world business applications. Presently, the Hy3 preview model has been integrated into multiple Tencent products, including Tencent Cloud, Yuanbao, ima, CodeBuddy, WorkBuddy, and QQ.