Same prompt.
Every model. Side-by-side.

Drop a prompt, pick models, hit run. We'll fan it out and stream responses head-to-head. Attach a harness to grade them automatically.