Tencent has officially released its open-source AI model Hy3. The model uses a Mixture-of-Experts (MoE) architecture with 295 billion total parameters. Of those, 21 billion are active at any given time, plus 3.8 billion parameters for an added MTP layer. It handles context lengths up to 256,000 tokens. Tencent claims Hy3 matches the performance of models two to five times its size, based on internal testing and blind evaluations. In a blind evaluation by 270 experts, Hy3 scored 2.67 out of 4, outperforming GLM-5.1 at 2.51. The hallucination rate dropped from 12.5 percent to 5.4 percent in internal testing. Hy3 is available under an Apache 2.0 license on Hugging Face, ModelScope, and GitHub. An FP8-quantized version is also available. Support for platforms like OpenRouter and Cline is planned. Tencent has already integrated the model into its own products, including WorkBuddy, Yuanbao, WeChat, and the game assistant for 'Path of Exile: Advent.'
Tencent's release of Hy3 highlights its ongoing efforts to expand its AI capabilities. The model's performance, as stated by the company, suggests a significant advancement in efficiency and effectiveness. The company's integration of Hy3 into its existing products indicates a strategic move to enhance user experience and functionality. The model's availability on multiple platforms also demonstrates Tencent's commitment to open-source collaboration.
Source: thedecoder
