ARC-AGI-2 leaderboard: who is #1

Name: ARC-AGI-2 rankings
Creator: ARC Prize
License: https://creativecommons.org/licenses/by/4.0/

ARC Prize · Score, higher is better · 78 models · published Sep 22, 2026

Top 25 of 78 models, best variant per model as published

#	Model	Score
1	GPT-6 Astraas “GPT-6 Astra (Max)”	95.0%
2	Claude Opus 5.5as “Claude Opus 5.5 (High)”	93.3%
3	GPT-5.6 Solas “GPT-5.6 Sol (Max)”	92.5%
4	Claude Opus 5as “Claude Opus 5 (Max)”	90.4%
5	Claude Fable 5.1as “Claude Fable 5.1 (Max)”	90.0%
6	Claude Fable 5as “Claude Fable 5 (Max)”	89.2%
7	GPT-5.5as “GPT-5.5 (XHigh)”	85.0%
8	Gemini 3.7 Flashas “Gemini 3.7 Flash (High)”	84.6%
9	Gemini 3as “Gemini 3 Deep Think (2/26)”	84.6%
9	GPT-5.5 Proas “GPT-5.5 Pro (High)”	84.6%
11	GPT-5.6 Terraas “GPT-5.6 Terra (Max)”	83.9%
12	GPT-5.4 Proas “GPT-5.4 Pro (XHigh)”	83.3%
13	Gemini 3.1 Pro Previewas “Gemini 3.1 Pro (Preview)”	77.1%
14	Dots3 Noteas “Dots3-Note Preview (Max)”	76.8%
15	Claude 4.7as “Claude 4.7 (Max)”	75.8%
16	GPT-5.4as “GPT-5.4 (XHigh)”	74.0%
17	Claude Opus 4.8as “Claude Opus 4.8 (High)”	72.1%
17	Gemini 3.5 Flashas “Gemini 3.5 Flash (High)”	72.1%
19	Claude Opus 4.6as “Claude Opus 4.6 (120K, High)”	69.2%
20	Grok 4.6as “Grok 4.6 (XHigh)”	67.1%
21	Grok 4.20as “Grok 4.20 (Reasoning)”	65.1%
22	DeepSeek V4 Flash Visionas “DeepSeek V4 Flash 0731 (Max)”	61.4%
23	DeepSeek V4 Pro 0813as “DeepSeek V4 Pro 0813 (Max)”	61.3%
24	Claude Sonnet 4.6as “Claude Sonnet 4.6 (High)”	60.4%
25	Gemini 3.6 Flashas “Gemini 3.6 Flash (High)”	60.4%