Artificial Analysis Intelligence Index

The Artificial Analysis Intelligence Index is like a report card that combines scores from several different subject tests into one overall grade. It runs a model through a range of distinct evaluations—math, coding, reading comprehension, and more—under the same conditions, then rolls the results into a single number showing how well the model performs overall.

A higher number means a better overall result, but there's a tendency for scores to improve the longer a model spends working on a problem. Because of this, the index is often paired with a chart plotting score against time taken. This lets you compare which model performs best given the same amount of time, or which is fastest at a given level of ability.

The organization behind this index independently scores models from many different companies, separate from the companies that actually build the models. So when this number shows up in an article, it points to who did the grading, not who made the model.

Articles use it to directly compare two models, as in: "Gemini 3.7 Flash scored 56 on high reasoning effort, up 4 points from 3.6 Flash's 52." It's also used to benchmark models from different companies against the same yardstick, as in: "It scored 61 on the Artificial Analysis Intelligence Index, putting it in the same tier as GPT-5.6 Sol."

A common misunderstanding is assuming this score is self-reported by the company that built the model. In fact, it's an independent evaluation produced by a separate benchmarking firm that runs the same tests across multiple models. A high score also doesn't mean superiority at every task—articles have noted cases where a model with a lower overall score outperformed higher-scoring models on specific tasks like spreadsheet analysis.