Arc
multimodal
+
+
+
+
About
ARC-AGI is a reasoning benchmark that tests fluid intelligence and the efficiency of skill acquisition on unknown tasks, measuring how quickly AI systems can learn new skills outside their training data. Designed to be easy for humans but challenging for AI, it focuses on core knowledge priors and genuine problem-solving ability without reliance on pre-existing knowledge. The benchmark evaluates generalization from limited information to new, unseen instances.
+
+
+
+
Evaluation Stats
Total Models1
Organizations1
Verified Results0
Self-Reported1
+
+
+
+
Benchmark Details
Max Score1
Language
en
+
+
+
+
Performance Overview
Score distribution and top performers
Score Distribution
1 models
Top Score
2.5%
Average Score
2.5%
High Performers (80%+)
0Top Organizations
#1Google
1 model
2.5%
+
+
+
+
Leaderboard
1 models ranked by performance on Arc
| License | Links | ||||
|---|---|---|---|---|---|
| Jun 17, 2025 | Creative Commons Attribution 4.0 License | 2.5% |
