The author tested nine local LLMs using an identical web‑development prompt on an RTX 3060 12GB GPU with 16 GB RAM running CachyOS via llama.cpp, spending roughly eight hours on the benchmark. The goal was to identify which model delivers frontier‑level performance while fitting within consumer‑grade hardware constraints. Results and rankings were shared for community rating.