Nonprofit AI research and benchmarks
ARC Prize Foundation
ARC Prize Foundation develops ARC-AGI benchmarks and runs competitions to advance open research into artificial general intelligence. The nonprofit measures how efficiently AI systems adapt to unfamiliar problems, focusing on capabilities that come naturally to humans. Researchers can evaluate systems and compete to publish open-source solutions. Its ARC-AGI-3 benchmark offers interactive, turn-based environments accessible through a browser or API, where agents must discover rules and goals, then apply what they learn across increasingly difficult levels.
Co-founders François Chollet and Mike Knoop serve on the board, with Greg Kamradt leading the foundation as president. Its research lineage begins with Chollet’s 2019 paper, On the Measure of Intelligence, which introduced ARC-AGI to investigate fluid intelligence rather than pattern memorization. Chollet and Knoop created the ARC Prize competition in 2024; the initiative expanded into a nonprofit foundation in 2025.
The foundation reports benchmark use by OpenAI, Anthropic, Google DeepMind and xAI. ARC Prize 2026 offers $2 million in prizes for open-source solutions to ARC-AGI-2 and ARC-AGI-3. As of August 2026, reported contributions included $2 million from Ndea and $1 million from xAI, supporting an independent nonprofit funded through donations.
1 talk
Newest first
1 speaker at AIE
Affiliations reflect their AIE appearances, not necessarily current employment.
Messages from the stage
Human ability defines the benchmark's scope
The audience discussion examined efficiency metrics and why tasks beyond human ability fall outside this benchmark's scope.
Affiliations reflect each recorded session, not necessarily current employment.