ARC-AGI-3
ARC Prize / Reasoning · Learns and generalizes from unfamiliar rules
What it measures and how
Fixes the standard harness, separate from provider-optimized adapters.
How this evidence is used
Under review: official Terms require written permission for systematic collection and aggregation. Official-result aggregation terms are unresolved. Standard harness and Provider Adapter results stay separate; the highest specialized-adapter score does not represent an ordinary model.
Evaluation limits & attribution
Under review: official Terms require written permission for systematic collection and aggregation. Official-result aggregation terms are unresolved. Standard harness and Provider Adapter results stay separate; the highest specialized-adapter score does not represent an ordinary model.
Data license: Leaderboard-data usage terms need confirmation.
Results are published by ARC Prize. Original scores and the NexGateHub consensus index use different scales and cannot be added directly.