I use it exclusively as a server for LLM inference - first more for research but have been using it for coding now that there are sufficiently capable models that run on. For that I’m very happy.