What is the CSI?

The Capability-Seconds™ Index measures how efficiently frontier AI models convert compute time and money into useful output. It combines three dimensions—quality, speed, and cost—into a single number that tracks the practical value of AI inference over time.

A higher CSI means models are getting better, faster, or cheaper—or all three. The index uses a natural-log scale, so movements are directly interpretable: every +0.69 points represents a 2× improvement in efficiency, and every +1.10 points represents a 3× improvement. A negative move of the same magnitude means the opposite.

See Methodology for the full formula and scoring details, or read the foundational thesis in The Copernicus Problem.