llama.cpp adds support for Tencent Hy 4 preview architecture
September 4, 2026
The llama.cpp repository has integrated support for the Tencent Hy 4 (hy_v4) architecture. This update allows for local inference of the Hy 4 model series via the ggml backend.
HOW THIS AFFECTS YOU
●
builderYou can now run Tencent Hy 4 models on local hardware using llama.cpp.
●
researcherThis provides an immediate inference pathway for evaluating the Hy 4 architecture's performance.