I think you have to distinguish between self-hosted to run on CPU (like LLAMA), on consumer GPU or on big GPUs. I find the market currently very confusing. I'm

I think you have to distinguish between self-hosted to run on CPU (like LLAMA), on consumer GPU or on big GPUs. I find the market currently very confusing. I'm especially interested since the data center I'm working for is sitting on a bunch of A100 and I get daily requests of people asking for LLMs tuned to specific cases, who can't or won't use OpenAI for various reasons.