BenchmarksReasoning
ARC-AGI-2
The ARC-AGI-2 abstract-reasoning puzzle benchmark (semi-private set), the harder static successor to ARC-AGI-1.
ReasoningaccuracyHigher is better
What this benchmark measures
The ARC-AGI-2 abstract-reasoning puzzle benchmark (semi-private set), the harder static successor to ARC-AGI-1.
Rows on this page are sourced from public benchmark artifacts, leaderboard exports, or source-linked model reports. Each row keeps benchmark version, source model name, and available run details attached to the score.
The metric shown here is accuracy. It should be interpreted within ARC-AGI-2, not compared as part of a site-wide ranking.
What to be careful about
Public and semi-private splits differ; keep the reported effort/compute as run context.
No composite ranking
evals.report never combines benchmarks. accuracy on ARC-AGI-2 is its own number — don’t average it with other metrics.