A ChangeEnable Research Project
Pick the model you use, the kind of queries you send, and how often. See what it adds up to in terms you can feel. These are estimates, not measurements — the disclosure gap is the whole point.
Why some labs are missing. Anthropic, Meta, xAI, and DeepSeek have not disclosed per-query figures, and we don't show estimates for them here because any number we put on screen would travel without context. The empty rows on the Scorecard are the point. Numbers above come from Google's Gemini technical paper (measured), Epoch AI (GPT-4o estimate), and the University of Rhode Island AI Lab (GPT-5 and o3 estimates). See the methodology for the full evidence base.