Pricing: We track blended costs (Input + Output) per 1 Million tokens using the OpenRouter API.
Latency: We run a standardized test suite of 50 prompts ranging from "Hello World" to complex Python reasoning tasks. We measure the Time to First Token (TTFT) and Total Inference Time.
Efficiency Score: A proprietary metric calculated as (1 / Blended Cost) * (1 / Latency). Higher score = Better value per dollar.