Tencent releases Hy3 open-source model that allegedly matches models up to five times its active size

Tencent has released Hy3, an open-source mixture-of-experts language model that utilizes 21 billion active parameters out of a 295-billion parameter total. The company claims this architecture allows the model to match the performance of much larger systems while reducing hallucination rates to 5.4 percent. By optimizing efficiency, the release highlights a broader industry trend toward achieving high-level AI performance with lower computational requirements.