ARC-AGI leaderboard
ARC-AGI
1 models tested · Updated 2025-07-09 · Verified sources only
Grok 4 leads at 66.6%
1
Grok 4
xAI · Blog/xAI · 2025-07-09
New SOTA on ARC-AGI v1 among closed models. Announced alongside Grok 4 launch.
66.6%
benchmark.space canonical public ledger of AI capability