A
AI ComputeCloud activity meter

METHODOLOGY

How we estimate energy use

We start with the tokens recorded on this computer, then compare them with published research on similar AI models. Because the exact systems used by OpenAI and Anthropic are private, the result is shown as a range.

Best estimate

1What we knowTokens and model names recorded by your apps
2Closest comparisonPublished tests of similar AI models
3Allow for unknownsReused information and data-centre differences
4Energy rangeLow, best and high estimates

TWO ESTIMATES

How the sources compare

Each source starts with the same tokens recorded on your computer, so the results can be compared fairly.

BY MODEL

Where the estimate comes from

UNCERTAINTY

What moves the range

MEASURED INPUT

Token evidence behind this date

How to read these numbers

The researchers could not test the exact private versions of Claude and GPT used for your requests. The best estimate is useful for comparing days, projects and trends. The wider range shows how much the real answer could change.