Wiki
Tags
About
FAQ
RSS
AIPOST

ARC-AGI

ARC-AGI is a benchmark run by ARC Prize that evaluates AI's abstract reasoning and ability to adapt to unfamiliar problems.

1 article
Last mentioned

ARC-AGI (Abstraction and Reasoning Corpus for Artificial General Intelligence) is a benchmark proposed by AI researcher François Chollet in 2019. It measures how well AI adapts to unfamiliar problems using tasks that are easy for people but hard for AI. The nonprofit ARC Prize runs the benchmark and has released new versions.

Earlier versions asked systems to infer rules from input and output examples of grid puzzles. The third version, ARC-AGI-3, places an AI in interactive game environments without instructions, where it must explore and work out the goals and rules, and it also compares the AI's efficiency with that of human players.

This entry is based on AIPOST articles and widely known facts. If something is wrong, please send us a correction request.

Articles covering this entry

GPT-6 Astra’s 99.9% and 62.7% scores on ARC-AGI-3 came from different evaluation setups.

Read article

Related entries

Entries that often appear in the same articles.

Browse the wikiReport & corrections

OpenAI

OpenAI is a US AI research and product company known for its GPT models and ChatGPT.

60 articles

benchmark

An AI benchmark is a set of tasks and criteria used to measure and compare models’ performance or behavior.

45 articles

GPT

GPT is OpenAI's family of large language models, which underpins ChatGPT and the OpenAI API.

25 articles
AIPOST
© 2026 AIPOST. All rights reserved.

AIPOST is an AI publication covering practical AI, AI security, performance, startups, health, ethics and industry news. No account is needed, and our privacy policy explains how we handle personal information.