Local models

Running a private LLM on your own hardware: what it actually takes

GPU memory, quantization, serving software, fine-tuning and evaluation: a practical guide to running open-weight language models on servers you control.