# ARC-AGI-2

> Visual grid puzzles that are easy for people and new to every model.

Source: https://www.superintelligence.com.au/benchmarks/arc-agi-2/
Data updated 2026-10-07. Benchmark data from Epoch AI (CC BY) and official leaderboards.

Best score today: 95% (+41 pts in 12 months).
Human comparison: The test’s creators had more than 400 members of the public attempt the puzzles, and every puzzle was solved by at least two of them. They published no single average score, so there is no one number to draw here.

## What it tests

Grid puzzles where you work out a hidden rule from a few examples, then apply it to a new grid.

## Why it matters

It tests adapting to something new, which is closer to general intelligence than a knowledge test.

## Caveats

Some top results use very large amounts of computing per task, which the headline score doesn’t show. Results come from the ARC Prize leaderboard rather than Epoch AI’s own runs.

## Best score by quarter

| Quarter | Best score | Record held by |
| --- | --- | --- |
| Q3 2024 | 1% | o1-mini (unknown thinking) |
| Q4 2024 | 1% | o1-mini (unknown thinking) |
| Q1 2025 | 3% | o3-mini (high) |
| Q2 2025 | 9% | Claude Opus 4 |
| Q3 2025 | 16% | Grok 4 (Thinking) |
| Q4 2025 | 54% | GPT-5.2 Pro |
| Q1 2026 | 85% | Gemini 3 Deep Think |
| Q2 2026 | 89% | Claude Fable 5 (max) |
| Q3 2026 | 95% | GPT-6 Astra (max) |
| Q4 2026 | 95% | GPT-6 Astra (max) |

## Top scores right now

| Rank | Model | Developer | Score | Released |
| --- | --- | --- | --- | --- |
| 1 | GPT-6 Astra (max) | OpenAI | 95% | 2026-09-03 |
| 2 | GPT-6.1 Sol (max) | OpenAI | 94% | 2026-09-29 |
| 3 | GPT-6 Astra (xhigh) | OpenAI | 93% | 2026-09-03 |
| 4 | Claude Opus 5.5 (high) | Anthropic | 93% | 2026-09-22 |
| 5 | GPT-5.6 Sol (max) | OpenAI | 93% | 2026-07-09 |
| 6 | Claude Opus 5.5 (xhigh) | Anthropic | 93% | 2026-09-22 |
| 7 | GPT-6 Astra (high) | OpenAI | 92% | 2026-09-03 |
| 8 | GPT-6 Astra (medium) | OpenAI | 92% | 2026-09-03 |
| 9 | Claude Opus 5.5 (max) | Anthropic | 92% | 2026-09-22 |
| 10 | GPT-6.1 Sol (high) | OpenAI | 92% | 2026-09-29 |
| 11 | GPT-6.1 Sol (xhigh) | OpenAI | 92% | 2026-09-29 |
| 12 | Claude Opus 5 (max) | Anthropic | 90% | 2026-07-24 |
| 13 | GPT-5.6 Sol (xhigh) | OpenAI | 90% | 2026-07-09 |
| 14 | Claude Fable 5.1 (max) | Anthropic | 90% | 2026-09-01 |
| 15 | Claude Fable 5.1 (xhigh) | Anthropic | 90% | 2026-09-01 |
| 16 | GPT-6 Sol (max) | OpenAI | 90% | 2026-09-22 |
| 17 | Claude Fable 5 (max) | Anthropic | 89% | 2026-06-09 |
| 18 | Gemini 3.8 Flash (high) | Google DeepMind | 89% | 2026-09-02 |
| 19 | Claude Fable 5.1 (high) | Anthropic | 89% | 2026-09-01 |
| 20 | Claude Fable 5 (xhigh) | Anthropic | 88% | 2026-06-09 |
| 21 | Claude Opus 5 (high) | Anthropic | 88% | 2026-07-24 |
| 22 | Claude Fable 5 (high) | Anthropic | 88% | 2026-06-09 |
| 23 | Claude Opus 5.5 (medium) | Anthropic | 88% | 2026-09-22 |
| 24 | GPT-6.1 Sol (medium) | OpenAI | 87% | 2026-09-29 |
| 25 | Claude Fable 5.1 (medium) | Anthropic | 86% | 2026-09-01 |
| 26 | GPT-5.6 Sol (high) | OpenAI | 85% | 2026-07-09 |
| 27 | GPT-6 Astra (low) | OpenAI | 85% | 2026-09-03 |
| 28 | GPT-5.5 (xhigh) | OpenAI | 85% | 2026-04-23 |
| 29 | Gemini 3.7 Flash (high) | Google DeepMind | 85% | 2026-08-13 |
| 30 | Gemini 3 Deep Think | Google DeepMind | 85% | 2026-02-12 |
| 31 | GPT-5.5 Pro (high) | OpenAI | 85% | 2026-04-23 |
| 32 | GPT-5.5 Pro (xhigh) | OpenAI | 84% | 2026-04-23 |
| 33 | GPT-5.6 Terra (max) | OpenAI | 84% | 2026-07-09 |
| 34 | GPT-5.4 Pro (xhigh) | OpenAI | 83% | 2026-03-05 |
| 35 | GPT-5.5 (high) | OpenAI | 83% | 2026-04-23 |
| 36 | Gemini 3.8 Flash (medium) | Google DeepMind | 83% | 2026-09-02 |
| 37 | Claude Fable 5 (medium) | Anthropic | 83% | 2026-06-09 |
| 38 | Claude Fable 5.1 (low) | Anthropic | 78% | 2026-09-01 |
| 39 | GPT-6 Sol (xhigh) | OpenAI | 78% | 2026-09-22 |
| 40 | Gemini 3.8 Flash (low) | Google DeepMind | 78% | 2026-09-02 |
| 41 | Gemini 3.1 Pro Preview | Google DeepMind | 77% | 2026-02-19 |
| 42 | Claude Fable 5 (low) | Anthropic | 77% | 2026-06-09 |
| 43 | GPT-6.1 Sol (low) | OpenAI | 77% | 2026-09-29 |
| 44 | Claude Opus 4.7 (max) | Anthropic | 76% | 2026-04-16 |
| 45 | GPT-5.6 Terra (xhigh) | OpenAI | 74% | 2026-07-09 |
| 46 | GPT-5.4 (xhigh) | OpenAI | 74% | 2026-03-05 |
| 47 | Gemini 3.5 Flash | Google DeepMind | 72% | 2026-05-19 |
| 48 | Claude Opus 4.8 | Anthropic | 72% | 2026-05-28 |
| 49 | GPT-5.5 (medium) | OpenAI | 70% | 2026-04-23 |
| 50 | Claude Opus 5.5 (low) | Anthropic | 70% | 2026-09-22 |

Download all data: https://www.superintelligence.com.au/data/benchmarks.csv