If you already have some big VRAM GPUs, open source LLM lets you utilize the computing power at no cost (well, you could argue there is TNB bill).