Tencent Urgently Expands Inference Capacity for Hunyuan Hy4 Preview on WorkBuddy

Tencent has urgently expanded the inference cluster capacity for its newly released Hunyuan Hy4 preview model on the WorkBuddy platform to address queuing issues caused by a surge in user demand. The model, launched on August 28, 2026, experienced high traffic immediately upon release due to its enhanced Agent capabilities. This surge highlights the strong demand for advanced, large-scale Mixture-of-Experts (MoE) models in productivity applications, while also illustrating the immense computational pressure that next-generation LLMs place on cloud infrastructure. The Hy4 preview model is a 770-billion parameter MoE model (with 49 billion active parameters and a 1M context length) that currently lacks multimodal capabilities, meaning WorkBuddy will automatically route image or video tasks to other models. To manage load, Tencent has implemented daily free usage limits until September 10, 2026, and extended the free trial of the older Hy3 model.