Benchmarks/ARC-AGI-2

ARC-AGI-2

ARC-AGI-2 — the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data.

reasoning

Models Tested

83.3

Top Score

14.7

Average Score

Rankings

#	Model	Score
1	GPT-5.4 ProOpenAI	83.3
2	Gemini 3.1 Pro PreviewGoogle DeepMind	77.1
3	GPT-5.4OpenAI	74.0
4	Claude Opus 4.6Anthropic	69.2
5	Claude Sonnet 4.6Anthropic	60.4
6	GPT-5.2 ChatOpenAI	52.9
7	GPT-5.2OpenAI	52.9
8	Claude Opus 4.5Anthropic	37.6
9	Gemini 3 Flash PreviewGoogle DeepMind	33.6
10	Gemini 3 ProGoogle	31.1
11	GPT-5 ProOpenAI	18.3
12	GPT-5.1OpenAI	17.6
13	GPT-5.1 ChatOpenAI	17.6
14	Grok 4xAI	16.0
15	Claude Sonnet 4.5Anthropic	13.6
16	M Kimi K2.5moonshotai	11.8
17	GPT-5 ChatOpenAI	9.9
18	GPT-5OpenAI	9.9
19	Claude Opus 4Anthropic	8.6
20	o3OpenAI	6.5
21	o4 MiniOpenAI	6.1
22	Claude Sonnet 4Anthropic	5.9
23	Grok 4 FastxAI	5.3
24	ZA GLM 5 Turboz-ai	4.9
25	ZA GLM 5z-ai	4.9
26	Gemini 2.5 ProGoogle DeepMind	4.9
27	o3 ProOpenAI	4.9
28	GPT-5 MiniOpenAI	4.4
29	DeepSeek V3.2DeepSeek	4.0
30	Claude Haiku 4.5Anthropic	4.0
31	o3 MiniOpenAI	3.0
32	GPT-5 NanoOpenAI	2.6
33	DeepSeek-R1DeepSeek	1.3
34	Qwen3-235B-A22B-Instruct (Jul 2025)Alibaba	1.3
35	DeepSeek-R1 (May 2025)DeepSeek	1.1
36	Claude 3.7 SonnetAnthropic	0.9
37	Claude 3.7 Sonnet (thinking)Anthropic	0.9
38	o1-miniOpenAI	0.8
39	GPT-4.5OpenAI	0.8
40	Gemini 1.5 Pro (Sept 2024)Google	0.8
41	Grok 3 MinixAI	0.4
42	GPT-4.1OpenAI	0.4
43	Grok 3 Mini BetaxAI	0.4
44	Grok 3xAI	0.1
45	GPT-4.1 MiniOpenAI	0.1
46	GPT-4.1 NanoOpenAI	0.1
47	Grok 3 BetaxAI	0.1
48	Llama 4 MaverickMeta	0.1
49	Llama 4 ScoutMeta	0.1
50	GPT-4o (2024-11-20)OpenAI	0.1
51	GPT-4o-mini (2024-07-18)OpenAI	0.1
52	GPT-4o-miniOpenAI	0.1

Related Benchmarks