One prompt sends your question to GPT-4o, Claude, Gemini, Llama, and others in parallel. You see every response. Then each model evaluates the others using anonymous labels (Response A, B, C), so they rank by quality, not brand. We aggregate those rankings and a chairman model synthesizes everything into one answer. So you get one place to read the outcome, plus full transparency: every response and every ranking is there if you want to dig in.