ARC-AGI-3
The first interactive reasoning benchmark designed to measure human-like intelligence in AI agents.
Links
What is ARC-AGI-3?
ARC-AGI-3 is an interactive reasoning benchmark which challenges AI agents to explore novel environments, acquire goals on the fly, build adaptable world models, and learn continuously.
A 100% score means AI agents can beat every game as efficiently as humans.
Instead of solving static puzzles, agents must learn from experience inside each environment—perceiving what matters, selecting actions, and adapting their strategy without relying on natural-language instructions.
How it measures intelligence
- 100% human-solvable environments
- Skill-acquisition efficiency over time
- Long-horizon planning with sparse feedback
- Experience-driven adaptation across multiple steps
As long as there is a gap between AI and human learning, we do not have AGI.
ARC-AGI-3 makes that gap measurable by testing intelligence across time, not just final answers—capturing planning horizons, memory compression, and the ability to update beliefs as new evidence appears.
Design principles
- Easy for humans to pick up quickly
- No pre-loaded knowledge or hidden prompts
- Clear goals + meaningful feedback
- Novelty that prevents brute-force memorization
Features
ARC-AGI-3 includes replayable runs, a developer toolkit for agent integration, and a UI designed for transparent evaluation.
Replays + Evaluation
Inspect agent behavior through preview replays—track decisions, actions, and reasoning in a structured timeline.
Browse a sample replayTools + UI
Integrate your agent using the ARC-AGI-3 toolkit, then use the interactive UI to test and iterate.
Play and testDocs
Everything you need to build agents: environments, API usage, and integration guidance.
Read the docs