Intelligence Index

Comparing AI models is messy, because every model is tested on a different pile of benchmarks and every lab picks the ones that flatter it. The Intelligence Index is the independent benchmarking firm Artificial Analysis trying to fix that: they run a fixed set of evaluations across models themselves and combine the results into one number, so you can line models up side by side.

Treat it as a rough guide rather than a verdict. A single score hides which tasks a model is good at, and a model that scores lower overall may still be the better choice for your particular job. It is most useful as a sanity check against marketing claims, and for spotting the trade-off between how smart a model is and how fast or cheap it runs.