Exactly. You can buy on server grade mobo and CPU, stock it with ddr4 and run qwen and basically get gpt 3+ level quality now. A year from now, I imagine we'll see even better efficiency.