Abstract reasoning · %
ARC-AGI-3
Interactive, game-like third version of ARC-AGI that tests exploration and learning inside novel environments.
Active Contamination risk: Low 3 results · 3 models Updated Oct 22, 2026 Newest result: 2026-10-22 Official leaderboard ↗
What it measures
Score on the semi-private set as published by ARC Prize.
Limitations
New benchmark; only the top entries are cited (terms).
Contamination risk: Low
Semi-private set.
Leaderboard
3 models
| # | Model | Provider | Score | Rank | Setting | Source type | Date |
|---|---|---|---|---|---|---|---|
| 1 | GPT-6 Astra - Provider Adapter (High) | OpenAI | 99.9% | 1 | Semi-private set; CoT | Official leaderboard | Oct 22, 2026 |
| 2 | Gemini 4 Argon - Provider Adapter (Low) | 97.5% | 6 | Semi-private set; CoT | Official leaderboard | Oct 22, 2026 | |
| 3 | GPT-6.1 Sol - Provider Adapter (XHigh) | OpenAI | 96.4% | 10 | Semi-private set; CoT | Official leaderboard | Oct 22, 2026 |
Data sources & licences
| Source | Licence / terms | How we reuse it | Last fetched |
|---|---|---|---|
| ARC Prize leaderboard ARC Prize Foundation, arcprize.org/leaderboard | ARC Prize Terms: no copying/aggregation for commercial purposes without permission | Restricted – only top entries cited top 10 per set only (terms) | Oct 22, 2026 |
FAQ
What does ARC-AGI-3 measure?
Interactive, game-like third version of ARC-AGI that tests exploration and learning inside novel environments. Score on the semi-private set as published by ARC Prize.
How reliable is ARC-AGI-3?
New benchmark; only the top entries are cited (terms). Contamination risk: Low – Semi-private set.