On Monday, Tencent launched Hunyuan Hy3, a new reasoning model that integrates both fast and slow thinking. Built on a Mixture of Experts (MoE) architecture, Hy3 features 295 billion total parameters with 21 billion activated parameters and supports a context window of up to 256K tokens. Hy3 has been integrated into several Tencent products and services, including WorkBuddy/CodeBuddy, Yuanbao, Marvis, and ima. Its API is now available through Tencent Cloud’s TokenHub platform.
Tencent has priced Hy3 at RMB 1 ($0.15) per million input tokens and RMB 4 ($0.59) per million output tokens. Cached input tokens are billed at RMB 0.25 ($0.037) per million tokens. [TechNode reporting]