llama.cpp adds support for Hy3 model via PR

llama.cpp has a pull request adding support for the Hy3 model, with GGUF quantizations available. Early testing shows coherent output at ~10-11 t/s on a 5090 with Q2_K quantization.