Intelligence Index vs. Cost per Task

Each point is one AI model. The best models are in the top left corner: a high score for a low cost.

No limit

Loading model data …

On the Pareto line Other models in the filter Hidden by the time filter

How to read this chart

The X axis shows the cost to run one task of the Artificial Analysis Intelligence Index test suite. The scale is logarithmic: each step to the right multiplies the cost by ten.

The Y axis shows one of three scores. Use the buttons in the filter row to pick one. A higher value is better on all of them.

The blue line is the Pareto line. A model is on this line when no other shown model is both smarter and cheaper. These models give you the best value for money.

The response time filter

Reasoning models can think for a long time before they answer. The "Max response time" slider hides slow models. The value is the median end-to-end response time in seconds. This is the total time from the request to the last token of a 500-token answer. When you move the slider, the Pareto line is computed again from the models that remain.

Point at a model to see its details. Use the Tab key to step through the models with the keyboard.