reasoning benchmark
ARC-AGI
The Abstraction and Reasoning Corpus for Artificial General Intelligence (ARC-AGI) is a benchmark designed to test general intelligence and abstract reasoning capabilities through visual grid-based transformation tasks. Each task consists of 2-5 demonstration pairs showing input grids transformed into output grids according to underlying rules, with test-takers required to infer these rules and apply them to novel test inputs. The benchmark uses colored grids (up to 30x30) with 10 discrete colors/symbols, designed to measure human-like general fluid intelligence and skill-acquisition efficiency with minimal prior knowledge.
Updated Aug 7, 2026
Published models7
Registry coverage7
MetricScore
EvidenceB
ARC-AGI leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | OP | 95.0% | 100.0% | 7 | C | |
| 2 | OP | 93.7% | 83.3% | 7 | C | |
| 3 | OP | 90.5% | 66.7% | 7 | C | |
| 4 | OP | 88.0% | 50.0% | 7 | C | |
| 5 | OP | 86.2% | 33.3% | 7 | C | |
| 6 | ME | 50.3% | 16.7% | 7 | C | |
| 7 | AC | 41.8% | 0.0% | 7 | C |
Score distribution
Top published rows on the benchmark's original scale.
ARC-AGI highlights
The top published results on this benchmark's own scale.
What is ARC-AGI?
Definition and scoring fields from the benchmark registry.
The Abstraction and Reasoning Corpus for Artificial General Intelligence (ARC-AGI) is a benchmark designed to test general intelligence and abstract reasoning capabilities through visual grid-based transformation tasks. Each task consists of 2-5 demonstration pairs showing input grids transformed into output grids according to underlying rules, with test-takers required to infer these rules and apply them to novel test inputs. The benchmark uses colored grids (up to 30x30) with 10 discrete colors/symbols, designed to measure human-like general fluid intelligence and skill-acquisition efficiency with minimal prior knowledge.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- ARC-AGI
- Modality
- image
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- arc-agi|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about ARC-AGI.
Which model scores highest on ARC-AGI?
GPT-5.5 is currently ranked first with 95.0%.
What does ARC-AGI measure?
The Abstraction and Reasoning Corpus for Artificial General Intelligence (ARC-AGI) is a benchmark designed to test general intelligence and abstract reasoning capabilities through visual grid-based transformation tasks. Each task consists of 2-5 demonstration pairs showing input grids transformed into output grids according to underlying rules, with test-takers required to infer these rules and apply them to novel test inputs. The benchmark uses colored grids (up to 30x30) with 10 discrete colors/symbols, designed to measure human-like general fluid intelligence and skill-acquisition efficiency with minimal prior knowledge.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
7 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.