Mini toolsARC-AGI-3
ARC-AGI-3: A New Challenge for Frontier Agentic Intelligence
An interactive benchmark for studying agentic intelligence through novel, abstract, turn-based environments in which agents must explore, infer goals, build internal models of environment dynamics, and plan effective action sequences without explicit instructions. A 100% score means AI agents can beat every game as efficiently as humans.
Agentic
Vision
Score
Max 100.0%
Released Apr 2026
3
Results
3
Models scored
7.80%
Top: GPT-5.6 Sol (Max)
0.80%
Median
Best results
Top primary scores; one row per model.
Frontier over time
Each dot is one model result; the line traces the running best score.
All data points share one date — no trend to plot.
All results
Showing one canonical row per model. Show all configurations
| # | Model | Score | Conditions | Eval date | Source | Flags |
|---|---|---|---|---|---|---|
| 1 | GPT-5.6 Sol (Max) | 7.80% | — | 09 Jul 2026 | Official leaderboard | Primary |
| 2 | GPT-5.6 Terra (Max) | 0.80% | — | 09 Jul 2026 | Official leaderboard | Primary |
| 3 | GPT-5.6 Luna (Max) | 0.20% | — | 09 Jul 2026 | Official leaderboard | Primary |
