tencent/Hy4-preview

Tencent releases Hy4 preview, a 770B parameter Mixture-of-Experts model with 49B active parameters. The card documents vLLM and SGLang deployment steps, confirming immediate usability for developers. It features Gated DSA attention and a native MTP layer for speculative decoding, positioning it as a high-performance open-source option.

README

tencent/Hy4-preview View on Hugging Face

Loading the README from Hugging Face…