ARC-AGI-2 is the second generation of the Abstraction and Reasoning Corpus published by the ARC Prize Foundation: grid puzzles meant to resist memorization and measure fluid reasoning, with public training and evaluation sets on GitHub (benchmark page). In this corpus it is the benchmark that a harness, not a new model, first took to 75% at under $8 a problem; two days later the foundation declared it saturated, and by February program synthesis had pushed it to 97.92%. The successor is ARC-AGI-3.
Benchmark
ARC-AGI-2
The ARC Prize Foundation's second abstraction-and-reasoning benchmark, whose climb from harness-driven 75% to declared saturation the corpus uses to date the end of fluid-intelligence tests as a ceiling.
Sources
Attribution and reporting linked by this record; links do not imply independent verification.
- Welcome to December 24, 2025 (theinnermostloop.substack.com)
Development timeline
Newest reporting dates first. Reporting dates may differ from event dates.
Cite or correct this record
Suggest a correction — opens a GitHub draft for you to review and submit.