ARC-AGI-3 Competition

The ARC-AGI-3 track challenges participants to build AI agents that interact with novel game environments, testing exploration, planning, memory, and goal-directed behavior.

ARC-AGI-3 is the first interactive reasoning benchmark. Unlike static tasks, AI agents must explore, adapt, and act within dynamic environments.

For general rules and the spirit of ARC Prize, see the ARC Prize 2026 overview.

Prizes - $850K Total

ARC Prize 2026 ARC-AGI-3 winners will be determined by participants' agent performance across the ARC-AGI-3 game environments.

Grand Prize (100%) - $700K

Awarded to the first eligible agent that scores 100% on the ARC-AGI-3 evaluation. If not won, the Grand Prize will continue during the next annual competition.

Top Score Award - $75K (guaranteed)

• 1st Prize: $40K
• 2nd Prize: $15K
• 3rd Prize: $10K
• 4th Prize: $5K
• 5th Prize: $5K

Milestone Prizes - $75K (guaranteed)

Participants who open source their solutions by the milestone deadlines are eligible for milestone prize money.

• Milestone #1 (June 30, 2026): 1st: $25K, 2nd: $10K, 3rd: $2.5K
• Milestone #2 (September 30, 2026): 1st: $25K, 2nd: $10K, 3rd: $2.5K

Submission Requirements

Submissions must be made through the designated Kaggle competition.

• No internet access during evaluation
• All code and methods must be open sourced to be eligible for prizes
• Hardware and compute limits will be announced with the competition launch

What ARC-AGI-3 Tests

ARC-AGI-3 evaluates five core capabilities:

1. Exploration - discovering rules through interaction
2. Percept, Plan, Action - observing, strategizing, and executing
3. Memory - retaining and applying learned information
4. Goal Acquisition - identifying objectives from context
5. Alignment - following implicit constraints

Learn more about the benchmark on the ARC-AGI-3 info page.

Previous Competitions

The ARC-AGI-3 Developer Preview was the first competition using the interactive reasoning format. See the ARC-AGI-3 Developer Preview learnings.