Used GPU plus local LLM cuts AI coding costs 93% in shared Ornith 9B setup

A setup pairing a used AMD Radeon RX 7900 XTX (24GB VRAM) bought for about $900 with the open-weight coding model Ornith-1.0 9B run locally via Ollama, said to cut AI coding subscription costs by roughly 93%, circulated widely online in July 2026.