CPU-only LLM inference benchmarked on $100 Celeron SBC

Tested 6 models from 0.6B to 8B on a Youyeetoo X1S (Celeron N5095, 16GB RAM) using Ollama CPU-only. Base config costs $100-130.