model selection

Hands-on, Bash-first guide to local AI on Linux: map hardware and use case; choose model families and sizes (Llama 3.1, Mistral, Qwen2, Phi-3) with the right quantization (Q4–Q6); pick a runtime (Ollama for simplicity, llama.cpp for control); install via apt/dnf/zypper; pull GGUFs; benchmark and iterate; includes RAM/VRAM rules, starter builds, and troubleshooting to get private, fast, predictable LLMs in ~15 minutes.