Benchmarks

Each benchmark compares the same prompt or brief across providers and models. Results may include HTML, code, timing or JSON.