Tencent Releases and Open-Sources Hunyuan Hy3 Preview with 295B Parameters

Gate News message, April 23 — Tencent unveiled and open-sourced Hunyuan Hy3 preview, a hybrid mixture-of-experts language model featuring fast and slow thinking fusion. The model comprises 295 billion total parameters with 21 billion active parameters, supporting a maximum context length of 256K tokens. Hy3 preview is the first model trained after Hunyuan’s restructuring and Tencent’s most intelligent model to date. The company reports significant improvements in complex reasoning, instruction following, in-context learning, code generation, agentic capabilities, and inference performance. Chief AI Scientist Yao Shunyu stated that Hy3 preview marks the first step of Hunyuan’s reconstruction and expressed hope that open-source community feedback will enhance the formal version’s practical utility.

Tencent is simultaneously scaling pretraining and reinforcement learning to elevate the model’s intelligence ceiling and collaborating with multiple Tencent products through deep co-design to improve real-world performance. The company is also beginning to explore specialized model capabilities.