ARC-AGI-3

Name: ARC-AGI-3
Keywords: ARC-AGI-3, AI benchmark, model evaluation, leaderboard, AI performance

AI benchmark results and model performance

Type: percentageReasoning

30.16%

30.16% - 30.16%

30.16% - Claude Opus 5

Scores Over Time

Individual benchmark scores plotted by date.

Organisation	Model	Reported	Top Score	Info	Self Reported	Source
Anthropic	Claude Opus 5	24 Jul 2026	30.16%	Relative Human Action Efficiency; ARC Prize Foundation verified semi-private evaluation; high effort	No	Source