Tencent Hunyuan has officially rolled out and open-sourced the preview version of its latest large language model, Hy4. This cutting-edge model is a powerhouse, boasting a staggering 770 billion parameters in total, with 49 billion actively engaged in processing. Its context length stretches beyond a remarkable 1 million, making it exceptionally adept at handling productivity tasks such as coding, office work, and scientific applications.

The model represents a significant leap forward in terms of scale, context length, and the breadth of data it can process. Through strategic collaboration with Tencent's internal products and the joint development of expert data sets, the intelligence level of Hy4 has been further refined and elevated. In Tencent's rigorous internal blind tests, the Hy4 preview version demonstrated a slight edge over its competitors, GLM 5.3 and Kimi K3, showcasing its superior performance.

Currently, the model has been seamlessly integrated into several Tencent products, offering users the opportunity to experience its capabilities firsthand through these platforms or access it via API. To gather valuable user feedback, related products will also provide a limited-time free trial period of two weeks, allowing users to explore and evaluate the model's potential.