Chinese tech giant Tencent unveiled the open-source AI model Hy3 on July 6, which outperforms OpenAI’s GPT-5.5. Hy3 features 295 billion total parameters, with 21 billion active parameters, and is designed to answer questions only when supported by evidence. According to Tencent’s benchmark results, Hy3 performs similarly to models like GLM-5.2 and DeepSeek-V4-Pro, and outperformed GPT-5.5 on the FrontierScience-Olympiad benchmark. Hy3 uses a Mixture-of-Experts (MoE) architecture and was trained with strict data-quality controls to reduce hallucinations and factual conflicts. Hy3 completed document-processing tasks using 47.4 percent fewer tokens than GLM-5.2. It has been released under the Apache 2.0 license for free commercial and research use.