Has anyone tried this with Llama 3.2 or any other “open source” options?