Skip to content

General model capability

ARC-AGI-2

Novel abstract visual-grid transformations.

About ARC-AGI-2

What it measures

Exact task accuracy at a reported inference-cost budget.

Scope & limitations

Public, semi-private and private evaluation sets differ. The Harbor 167-task package is not the full official private evaluation.

Evaluation environment

Reasoning system using the official task protocol.

Source

Original benchmark

Dataset & task catalog ↗