Every plotted model: composite intelligence (Y) against output price per million tokens (X, log). Dot size tracks context window; color is provider. The line traces the Pareto frontier — best intelligence at each price.

Overlay 2–3 models across six capability axes. Toggle models below.

Models × capabilities, colored red→amber→green. Click a column header to sort. Click a row for the full detail drawer.

Models ranked by agentic composite (LiveBench Agentic Coding + SWE-Bench), with cost-per-agentic-task and a Terminal-Bench leaderboard.