We are running known models like LLama, StableDiffusion etc. So pretty much the same :)
We are not planning on building our own models yet.