LLM Intelligence vs Cost

Top 200 models by Artificial Analysis Intelligence Index (v4.3), plotted against the cost to complete one intelligence-index task. Data pulled 2026-10-05 from artificialanalysis.ai. This is a snapshot, not a live benchmark.

0

The X axis shows cost per task (USD), with log scale on by default. Use the toggle to compare log and linear views. Hover a point for details; click to highlight the row in the table. Models without published cost-per-task data are listed in the table but not plotted. The green line is the best-in-class frontier: each diagonal segment connects successive models where no cheaper model scores higher.

Best in class for its price

A model makes this list when no cheaper model scores higher (ties: the cheaper one wins). These are the only models that are ever the smartest choice at some budget.

Methodology. Intelligence Index v4.3 and cost-per-task figures are published by Artificial Analysis and reproduced here with attribution. Cost per task is the USD price to complete their standard intelligence-index task set at a 3:1 input-output ratio. Where Artificial Analysis has not published a cost, the model appears in the table only. Scores from different index versions are not comparable; this snapshot uses one version throughout.

Best in class. The green diagonal line connects models where no cheaper model scores higher, tracing the top-left edge of the point cloud; anything below the line is dominated on intelligence per dollar at its own price point. License labels reflect weights availability at the data-pull date (open-weights releases after that date are not reflected); Kimi K3 ships under its own license with commercial conditions and is counted as open source.