llama.cpp

3f3b9a22 - model : support qwen3 rerank and embeddings