ARC-AGI-3 B
AI benchmark · Reasoning
What is ARC-AGI-3?
ARC-AGI-3 measures whether AI agents can explore novel interactive game environments, acquire goals, build world models and learn skills as efficiently as humans.
General capability mostly
“The first interactive reasoning benchmark designed to measure human-like intelligence in AI agents.”
arcprize.org/arc-agi/3 ↗“A 100% score means AI agents can beat every game as efficiently as humans.”
arcprize.org/arc-agi/3 ↗
Lineage
Benchmarks this one is built on and built from: versions, subsets, extensions and the indexes that combine them.
Built on
ARC-AGI-2 earlier version
Derived benchmarks
ARC-AGI-3 Provider Adapter extension
Tier B
Established: 15 or more models scored, or curated RL directories rate it.
How much stands behind a tier: usage, who backs it, a confirmed line from its own site or paper, and a clear read on X. High with three of the four, medium with two, low otherwise.
- 25 models scored
- in curated RL directories (2 pts)
On X · mixed, moderate buzz, from 14 posts about ARC-AGI-3 in the last month
Labs and competitors still report ARC-AGI-3 scores, but Astra and Gemini near 99% with provider harnesses suggest saturation, and its scoring method draws criticism.
How we got to tier B
- Adoption: 25 models scored. Tier B.
- On X: mixed, moderate buzz. “Labs and competitors still report ARC-AGI-3 scores, but Astra and Gemini near 99% with provider harnesses suggest saturation, and its scoring method draws criticism.” No move.
- In curated RL directories: 2 points, a floor of C.
- Confirmed: a line from its own site or paper places it.
A clear view on X moves a tier one step up or down; most entries don't move. How tiers work.
Where ARC-AGI-3 sits
Its domain on the map, live. Hover a logo for its name; click to open it. Open it on the full map
Next: browse all benchmarks. Entry from the RL Research daily scrape of public sources, 2026-11-05 snapshot.