Tencent released Hy4 Preview, an open-sourced 770-billion parameter Mixture-of-Experts (MoE) flagship model. The architecture utilizes 49 billion active parameters per token. It supports a context window exceeding 1 million tokens.
Tencent designed the model for complex tasks in software engineering, game development, and scientific research. Internal blind tests indicate Hy4 performance slightly surpasses Moonshot AI’s Kimi K3 and GLM 5.3. Developers can access the model through Tencent Cloud’s TokenHub and OpenRouter.