ARC-AGI
by Chollet / ARC Prize Foundation · open-source · Last verified 2026-03-17
ARC-AGI (Abstraction and Reasoning Corpus for Artificial General Intelligence) measures fluid intelligence through visual grid transformation puzzles. Models must infer transformation rules from three or fewer examples and apply them to a test grid — a task trivially solved by humans but historically extremely difficult for AI systems.
https://arcprize.org ↗B+
B+—Good
Adoption: AQuality: A+Freshness: ACitations: AEngagement: F
Specifications
- License
- Apache-2.0
- Pricing
- open-source
- Capabilities
- evaluation, abstract-reasoning, few-shot-generalization
- Integrations
- Use Cases
- model-evaluation, agi-research, reasoning-benchmarks
- API Available
- No
- Evaluated Models
- gpt-4o, claude-opus-4, gemini-2-5-pro, o3
- Metrics
- accuracy
- Methodology
- 400 public + 100 semi-private + 500 private tasks. Each task provides 3 input-output grid examples; model must predict the output for a test input. Evaluated with 2 submission attempts; accuracy = % tasks solved. Human baseline ≈ 85%.
- Last Run
- 2026-03-01
- Tags
- agi, abstract-reasoning, visual-patterns, few-shot, core-knowledge
- Added
- 2026-03-17
- Completeness
- 100%
Index Score
74.1Adoption
84
Quality
95
Freshness
88
Citations
86
Engagement
0