A common measure
of AI efficiency
How far does your AI go? We’re developing an independent way to measure the useful work AI tools deliver against the resources they consume.
Efficiency deserves
a clearer standard.
AI tools can complete the same task in very different ways. Some use fewer tokens. Others move work into hidden processing. Our aim is to make the full tradeoff visible, so efficiency can be judged alongside the quality of the result.
Compare the same work.
Evaluate tools against defined tasks and clear success criteria. Measure what was accomplished as well as the resources required.
Find the avoidable waste.
Examine repeated context, unnecessary model calls, and work that conventional software could handle more efficiently.
Make the evidence usable.
Develop understandable ratings and transparent reporting that help people choose and improve AI tools with confidence.
Useful work first.
Lower consumption is valuable when the result still meets the task. Efficiency must preserve quality, reliability, and safety.