Back to BenchmarksGoogle
👁️
ARC-AGI-2
Abstract Reasoning Corpus v2
Fluid intelligence benchmark; humans solve easily, AI struggles
👁️ Multimodal
Best: 97.9%
6 Models Evaluated
Champion
Gemini 3 Pro
84.6%
Best Score
Model Rankings
All models ranked by score on ARC-AGI-2. Scores are relative to the best score of 97.9%.
| # | Model | Provider | Score | vs Best |
|---|---|---|---|---|
| 1 | Gemini 3 Pro Elo 1486 | 84.6% | Best | |
| 2 | GPT-5.4 Pro Elo 1480 | OpenAI | 83.3% | -14.6% |
| 3 | GPT-5.5 Elo 1478 | OpenAI | 75% | -22.9% |
| 4 | Claude Opus 4.7 Elo 1500 | Anthropic | 69% | -28.9% |
| 5 | Claude Opus 4.6 Elo 1503 | Anthropic | 68.8% | -29.1% |
| 6 | Claude Sonnet 4.6 Elo 1460 | Anthropic | 58.3% | -39.6% |
Score Distribution
Gemini 3 Pro
84.6%
GPT-5.4 Pro
83.3%
GPT-5.5
75%
Claude Opus 4.7
69%
Claude Opus 4.6
68.8%
Claude Sonnet 4.6
58.3%
Run This Benchmark
Learn how to evaluate your own models on ARC-AGI-2.
ARC-AGI-2 is a standardized benchmark for evaluating AI model capabilities.
To run this benchmark on your own models, visit the official benchmark page or leaderboard for detailed instructions, code, and evaluation scripts.
Data compiled from official model cards, public leaderboards, and independent evaluations.