llama.cpp

40a961db - args: default --model to models/ + filename from --model-url or --hf-file (or else legacy models/7B/ggml-model-f16.gguf)