Artificial Analysis

Artificial Analysis is a company that ranks AI models by testing them. It was started by two people in Australia in 2023, and is now based in San Francisco.

What the company does isn't build models—it grades them. It runs models through a battery of tests, combines the results into a single score, and also measures how fast they are and how much they cost. The "index score of X" you see quoted every time a new model launches usually comes from this company.

So when this name shows up in an article, don't get the direction confused. A high score doesn't mean this company built the model. It's the referee, not a player. Also, since the index is a combination of multiple tests rather than a single one, you need to check which specific tests a model did well on to get the full picture.

The source behind most "intelligence index score of X" figures in our articles is this company. Grok 4.6's score of 61 in August 2026 is one such example. When a performance debate breaks out on a new model's launch day, both sides often show up waving this company's charts—it functions almost like the industry's shared scorecard.

"Ranked #1 on the Artificial Analysis Index" — this means it led in the combined score across multiple evaluations, not that it topped every single category.