Repository navigation
Windows x64 build of llama-server.exe and the other tools with the CUDA 12 backend, built from eb039c2 (build run). Each zip holds the executables, all the DLLs they load and the CUDA runtime DLLs.
llama-eb039c2-bin-win-cuda-12.4-x64.zip: GTX 900 series to RTX 40 series, NVIDIA driver R550 or newerllama-eb039c2-bin-win-cuda-12.8-x64.zip: RTX 50 series, NVIDIA driver R570 or newer
Unzip into one folder and run:
llama-server.exe -m SenseNova-Loop-9B-Preview-Q4_K_M.gguf -ngl 99 -c 16384 --port 8080
Supported GGUF files and details: docs/sensenova-loop.md