ARC Prize
Compare abstract reasoning, interactive learning and run costs, separated by benchmark version and framework.
Source & report coverage
- Source type
- Independent reasoning benchmarks
- Report snapshots
- 8 reports
- Data as of
- 2026-10-28
Sources and individual reports are listed separately. Each snapshot retains its subjects, tasks, metrics and conditions. External scores are not our own measurements and are not added across reports.
Reports & historical snapshots
- ARC-AGI-1 · Abstract reasoning →
Observed2026-10-28 · 291subject / run records ·2independent metrics
- ARC-AGI-2 · Abstract reasoning →
Observed2026-10-28 · 291subject / run records ·2independent metrics
- ARC-AGI-3 · Provider Adapter →
Observed2026-10-28 · 42subject / run records ·2independent metrics
- ARC-AGI-3 · standard framework →
Observed2026-10-28 · 68subject / run records ·2independent metrics
- ARC-AGI-1 · Abstract reasoning →
Observed2026-10-19 · 271subject / run records ·2independent metrics
- ARC-AGI-2 · Abstract reasoning →
Observed2026-10-19 · 272subject / run records ·2independent metrics
- ARC-AGI-3 · Provider Adapter →
Observed2026-10-19 · 38subject / run records ·2independent metrics
- ARC-AGI-3 · standard framework →
Observed2026-10-19 · 64subject / run records ·2independent metrics