Oracle has published a two-part guide on deploying the Private Large Language Model Service from the Private AI Services Container 26.2.1 on a CPU-powered OCI virtual machine. Part 1 covers creating the VM, downloading the large AVX-512 container image, and using the default LLM for chat. Part 2 will address downloading open-weight LLMs from Hugging Face with llama.cpp and vLLM runtimes, checking LLM state, and using Hugging Face tokens and SSL. The service can run on-premises, in an air-gap environment, or in a public cloud, with this tutorial focusing on CPU deployment in OCI. LLMs are resource-intensive, with larger parameter counts generally increasing resource requirements and slowing performance. Models are grouped into three size ranges: up to 10 billion parameters for simpler tasks like chat and RAG; 10 to 70 billion for medium complexity tasks including coding agents; and over 70 billion, such as GLM 5.2 and Ornith-1.0-397B, for challenging reasoning tasks. The service supports all three ranges on powerful Linux servers, with CPU memory (RAM) roughly equaling model size, and CPU cores providing scalable performance for parallel matrix operations.