Tencent expands inference clusters due to Hy4 preview demand

Tencent has announced an emergency expansion of its inference clusters following a surge in demand for its new large language model, Hy4 preview. Since the model was integrated into the WorkBuddy AI service on August 28, high volumes of requests from users and developers have resulted in task queues and waiting periods.

The Hy4 preview model features 770 billion total parameters, 49 billion active parameters, and a 1 million token context window. Due to limitations in high-end computing resources and concentrated peak-time concurrency, Tencent noted that delays may persist during certain periods.

To manage the load, Tencent is dynamically allocating resources and has suggested that users switch to the Hy3 model or use the service during off-peak evening hours to avoid queues. The free trial period for Hy4 preview is scheduled to run until September 10, with daily usage limits applied per user.