Reasoning and knowledge
ARC-AGI-2
3 models with a published score, best first. Higher is better.
95.0%GPT-6 AstraRank 1, published Oct 19, 2026
Leaderboard
Each model's latest published score. Every score links to its source.
| # | Model | ARC-AGI-2 | Date | Note | Source |
|---|---|---|---|---|---|
| 1 | O GPT-6 AstraRecordOpenAI | 95.0% | Oct 19, 2026 | Max setting | benchlm.ai |
| 2 | O GPT-6.1 SolOpenAI | 94.2% | Oct 19, 2026 | Max setting | benchlm.ai |
| 3 | A Claude Opus 5.5Anthropic | 93.3% | Oct 19, 2026 | High setting | benchlm.ai |
Record history
The best score ever published on ARC-AGI-2, and who held it. Each step is a new record.
| Date | Record holder | ARC-AGI-2 | Gain | Note |
|---|---|---|---|---|
| Oct 19, 2026 | GPT-6 AstraCurrent | 95.0% | First | Max setting |
Method
Semi-private set, via the BenchLM mirror. Higher is better. Board source: arcprize.org.
We collect the scores the board and the labs publish and run none ourselves. Ranks use each model's latest reading. 3 readings stored for this board, the newest from Oct 19, 2026.