The Hy3 model, formerly known as Hunyuan, utilizes a hybrid Mixture-of-Experts architecture featuring 295 billion total parameters. Tencent claims the model performs on par with flagship competitors while maintaining higher efficiency in reasoning, code generation, and agentic tasks. Beyond WorkBuddy, the model is now accessible through Tencent Design Miora and the Tencent Cloud TokenHub, which allows enterprises to manage multiple AI models via a single API.

Early performance metrics suggest rapid adoption. Within one week of its initial launch, Hy3 recorded a 68-fold increase in API calls compared to its predecessor and secured the top spot on the OpenRouter usage leaderboard. Poshu Yeung, Senior Vice President of Tencent Cloud International, stated that the company is focused on moving enterprises past experimentation toward scalable, practical deployment. By providing the model under an Apache 2.0 license and integrating it with regional platforms like Cafe24 and Metelix, Tencent aims to cement its position in the global AI infrastructure market.