Explore model capabilities in 3D.

Pareto frontier: No other configuration is at least as good on Terminal-Bench, speed, Intelligence Index, knowledge and price, and better on at least one. This multidimensional frontier is not a line on the 2D charts.

Speed: Median output speed over the 72 hours preceding the snapshot, measured by AA with approximately 10,000 input tokens. Intelligence Index v4.3 is a composite benchmark score, not human IQ; it already includes Terminal-Bench and Omniscience.

Knowledge: Choose the overall AA-Omniscience Index (All sciences covers all six domains, including humanities, law, business and software engineering) or the Software Engineering (SWE) domain index. Both range from −100 to 100: correct answers increase the score, errors lower it, and abstentions incur no penalty. A negative score means more errors than correct answers. Overall test accuracy is shown separately in the model card. Coordinates, clouds and the five-dimensional Pareto frontier update for the selected domain. SWE scores are available for all 146 configurations in this snapshot.

Price limit: In 3D, translucent clouds surround models at or below the selected price limit. Nearby regions merge smoothly; distances account for the logarithmic speed axis. Clouds show model neighborhoods, not exact prices at every position: more expensive models may also fall inside them. In 2D, the contour estimates the budget boundary between nearby measured models. Select a point for its exact price. All models remain visible. The default limit is $6 per million output tokens.

Price: USD per million output tokens, as reported by AA. Input costs and the number of reasoning tokens used are not compared here. A zero price means the source reports $0. GLM-5.3 FlashX is not included because this snapshot has no comparable AA measurements.