AGI progress has stalled.
New ideas are needed.
Details
ARC-AGI measures progress towards artificial general intelligence using a private evaluation dataset. ARC Prize 2024 is scored against this leaderboard and the Grand Prize is unlocked by any team that reaches at least 85%.
ARC Prize 2024 imposes intentional constraints: no internet access (so no GPT-4, Claude, etc.) to reduce leaking the test and Kaggle compute is fixed to target efficiency.
New high scores will be announced on the @arcprize X account.
All submissions must be made on Kaggle.
ARC-AGI-PUB 2024 High Scores
Ryan Greenblatt |
42% | 43% |
| icecuber 2020 |
39% | 17% |
| o1-preview |
21% | 18% |
| Claude 3.5 |
21% | 14% |
| o1-mini |
13% | 9.5% |
| GPT-4o |
9% | 5% |
| Gemini 1.5 |
8% | 4.5% |
Details
ARC-AGI-Pub is a secondary leaderboard measuring the ARC-AGI public evaluation set. No competitions or prizes are associated with this leaderboard. Its purpose to explore frontier model performance on ARC-AGI.
This leaderboard imposes no limitations on internet access and few limitations on compute. That means you can utilize state-of-the-art models.
Selected high score claims are verified using a semi-private evaluation set.
New high scores will be announced on the @arcprize X account.
Submit ARC-AGI-Pub high score claims here.
