Published 25 Mar 2026

Announcing ARC-AGI-3

A New Challenge for Frontier Agentic Intelligence

Today we're excited to announce the release of ARC-AGI-3, a series of hundreds of interactive environments and thousands of game-style levels that measure the frontier of agentic intelligence. ARC Prize 2026 is now live with over $2 million in prize. Compete to open source solutions to ARC-AGI-2 and ARC-AGI-3!

ARC-AGI-3 Launch Event

ARC-AGI-3 is the first fully interactive benchmark in the ARC-AGI series. ARC-AGI-3 represents hundreds of original turn-based environments, each handcrafted by a team of human game designers. There are no instructions, no rules, and no stated goals. To succeed, an AI agent must explore each environment on its own, figure out how it works, discover what winning looks like, and carry what it learns forward across increasingly difficult levels.

Humans score 100%. Frontier AI scores 0.26%.

Previous ARC-AGI benchmarks predicted and tracked major AI breakthroughs, from reasoning models to coding agents. ARC-AGI-3 points to what's next: the gap between AI that can follow instructions and AI that can genuinely explore, learn, and adapt in unfamiliar situations.

Are You Up for the Challenge?

Play the games - Play ARC-AGI-3 games in your browser or via API.

ARC Prize 2026 - $2,000,000 in prizes - Two competitions, two benchmarks, one goal: close the gap between human and artificial intelligence through open-source research. The ARC-AGI-3 Competition is a new kind of Kaggle competition where you build agents that play ARC-AGI-3 games. The ARC-AGI-2 Grand Prize honors the original benchmark format - this year, the grand prize is guaranteed and will be awarded to the best open-source solution. Learn more.

For researchers - Read the ARC-AGI-3 technical paper on the design and methodology behind ARC-AGI-3. Get started building locally or via api.

ARC-AGI-3 Technical Paper PreviewARC-AGI-3 Technical Paper

Launch event - shared publicly live on March 25 in San Francisco at Y Combinator HQ, featuring a fireside conversation between François Chollet (creator, ARC-AGI) and Sam Altman (CEO, OpenAI) on measuring intelligence on the path to AGI. Recording and details to be shared on our events page.

Let the Games Begin!

Play the GamesWin PrizesRead the Technical Paper

Good luck to everyone participating in ARC Prize 2026!