ARC-AGI-2 Cost Efficiency
X = ARC Prize reported cost/task (log). Y = ARC-AGI-2 %. Each line connects one model's published reasoning-effort levels, so the score-vs-cost tradeoff is visible per model. Shows each model line's current and previous model. Color = provider.
ARC-AGI-2 Cost Efficiency
ARC-AGI-2 Cost Efficiency
X = ARC Prize reported cost/task (log). Y = ARC-AGI-2 %. Each line connects one model's published reasoning-effort levels, so the score-vs-cost tradeoff is visible per model. Shows each model line's current and previous model. Color = provider.
Controls:
How to read this chart
Each line is one model run at several reasoning-effort levels on ARC-AGI-2. Every point shows the score and cost per task at that effort, with cheaper runs on the left. Up and to the left is better; a line that stays flat while moving right is buying little extra score with the extra spend. To keep the chart readable, each model line (e.g. Opus, Sonnet, Flash) shows only its current and previous model.
Data sources
More / ExperimentalARC-AGI-2 vs Cost/TaskX = ARC Prize reported cost/task (log). Y = ARC-AGI-2 %. Color = provider.Data: ARC PrizeOpen chartIQ BenchmarksARC-AGI-2 ScoresEach model's ARC-AGI-2 score. Color = provider.Data: ARC PrizeOpen chartCostTask EfficiencyEach dot shows the inverse of the effective-cost usage multiplier. Higher means less price-adjusted task work: 2× is about half the median task effort. Source-backed multipliers are preferred; lineage, peer, and 1× fallbacks are labeled in tooltips.Data: Artificial Analysis, ARC Prize, Vals.aiOpen chartCostIQ vs Effective CostEach model's estimated IQ plotted against effective cost per 1M I/O Tokens (sticker price × measured or imputed usage multiplier).Data: AI IQ methodologyOpen chartMore / ExperimentalTerminal-Bench 2.0 Benchmark ScoresLegacy Terminal-Bench 2.0 scores retained for historical comparison. Color = provider.Data: Terminal-BenchOpen chartMore / ExperimentalBenchmark Correlation HeatmapA pairwise matrix of source-backed benchmark score correlations across public models. Color shows Pearson r; each cell uses models with both benchmark scores.Data: ARC Prize, Epoch AI FrontierMath, Vals.ai +28 moreOpen chart