Which model leads on ARC-AGI?

Claude Fable 5 from Anthropic leads ARC-AGI with a score of 98.5. The median score across 71 tested models is 65.3.

Is ARC-AGI saturated?

Yes · the top model on ARC-AGI has reached 98.5 out of 100, within 5% of the theoretical ceiling. This benchmark is approaching saturation and may be replaced by a harder successor.

Does ARC-AGI predict performance on other benchmarks?

Yes · ARC-AGI scores correlate 1.00 with Artificial Analysis · Terminal-Bench Hard across 5 shared models. Models that do well on ARC-AGI tend to do well on Artificial Analysis · Terminal-Bench Hard.

How often is ARC-AGI data refreshed?

BenchGecko pulls updates daily. New model scores on ARC-AGI appear as soon as they are published by Epoch AI or the model provider.

Benchmark · ReasoningSettled

ARC-AGI

Name: ARC-AGI Benchmark
Creator: BenchGecko
License: https://creativecommons.org/licenses/by/4.0/

ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.

Updated 2026-09-22

Models tested

Top score

98.5

Claude Fable 5

Median

65.3

min 0.5

Top-5 spread

σ 0.4

Settled

The Frontier

Best score over time · one chart, every benchmark

Chart type

Frontier on ARC-AGI rose from 4.5 to 98.5 in 19 months · +94.0 points · latest leader Claude Fable 5 from Anthropic.

Pink dots = frontier records · 13 totalClick to open model page

Full rankings

71 models tested · sorted by score

#	Model	Score	Price
1	Claude Fable 5· Anthropic	98.5	$10.00
2	GPT-6 Astra· OpenAI	98.5	$10.00
3	Gemini 3.1 Pro Preview· Google DeepMind	98.0	$2.00
4	Claude Fable 5.1· Anthropic	97.5	$10.00
5	Claude Opus 5· Anthropic	97.5	$5.00
6	Claude Opus 5.5· Anthropic	97.5	$4.00
7	GPT-5.6 Sol· OpenAI	97.5	$2.00
8	GPT-5.6 Terra· OpenAI	96.5	$2.00
9	Gemini 3.7 Flash· Google DeepMind	95.5	$0.75
10	GPT-5.5· OpenAI	95.0	$5.00
11	GPT-5.4 Pro· OpenAI	94.5	$30.00
12	Kimi K3· moonshotai	94.5	$0.67
13	Claude Opus 4.6· Anthropic	94.0	$5.00
14	GPT-5.4· OpenAI	93.7	$2.50
15	Claude Opus 4.7· Anthropic	93.5	$5.00
16	Claude Opus 4.8· Anthropic	92.5	$5.00
17	Gemini 3.5 Flash· Google DeepMind	92.5	$1.50
18	Gemini 3.6 Flash· Google DeepMind	91.2	$0.75
19	DeepSeek V4 Pro 0813· DeepSeek	90.5	$0.55
20	GPT-5.2 Pro· OpenAI	90.5	$21.00
21	Grok 4.20· xAI	89.5	$1.25
22	DeepSeek V4 Flash 0731· DeepSeek	89.0	$0.02
23	GPT-5.6 Luna· OpenAI	88.0	$0.20
24	Grok 4.6· xAI	87.5	$2.00
25	Grok 4.5· xAI	87.2	$2.00
26	Claude Sonnet 4.6· Anthropic	86.5	$3.00
27	GPT-5.2· OpenAI	86.2	$1.75
28	Gemini 3 Flash Preview· Google DeepMind	84.7	$0.50
29	Claude Opus 4.5· Anthropic	80.0	$5.00
30	GLM 5.2· z-ai	77.0	$1.00
31	Gemini 3 Pro· Google DeepMind	75.0	n/a
32	GPT-5.1· OpenAI	72.8	$1.25
33	GPT-5 Pro· OpenAI	70.2	$15.00
34	Grok 4· xAI	66.7	n/a
35	GPT-5· OpenAI	65.7	$1.25
36	Kimi K2.5· moonshotai	65.3	$0.45
37	Claude Sonnet 4.5· Anthropic	63.7	$3.00
38	GPT-5.4 Mini· OpenAI	63.7	$0.75
39	MiniMax M2.5· minimax	63.7	$0.27
40	o3· OpenAI	60.8	$2.00
41	o3 Pro· OpenAI	59.3	$20.00
42	o4 Mini· OpenAI	58.7	$1.10
43	DeepSeek V3.2· DeepSeek	57.0	$0.28
44	GPT-5 Mini· OpenAI	54.3	$0.25
45	Gemini 3.5 Flash Lite· Google DeepMind	53.5	$0.30
46	GPT-5.4 Nano· OpenAI	51.5	$0.20
47	Grok 4 Fast· xAI	48.5	n/a
48	Claude Haiku 4.5· Anthropic	47.7	$1.00
49	GLM 5· z-ai	44.7	$0.60
50	Gemini 2.5 Pro· Google DeepMind	41.0	$1.25
51	Claude Sonnet 4· Anthropic	40.0	$3.00
52	Claude Opus 4· Anthropic	35.7	$15.00
53	o3 Mini· OpenAI	34.5	$1.10
54	Gemini 2.5 Flash· Google DeepMind	32.3	$0.30
55	o1· OpenAI	30.7	$15.00
56	Claude 3.7 Sonnet· Anthropic	28.6	n/a
57	R1 0528· DeepSeek	21.2	$0.50
58	GPT-5 Nano· OpenAI	20.7	$0.05
59	o1-preview· OpenAI	18.0	n/a
60	Grok-3 mini· xAI	16.5	n/a
61	R1· DeepSeek	15.8	$0.70
62	o1-mini· OpenAI	14.0	n/a
63	Qwen3 235B A22B Instruct 2507· Alibaba Qwen	11.0	$0.09
64	GPT-4.5· OpenAI	10.3	n/a
65	GPT-4.1· OpenAI	5.5	$2.00
66	Grok 3· xAI	5.5	n/a
67	U Magistral Small 1.1· Unknown	5.0	n/a
68	GPT-4o (2024-11-20)· OpenAI	4.5	$2.50
69	Llama 4 Maverick· Meta	4.4	$0.19
70	GPT-4.1 Mini· OpenAI	3.5	$0.40
71	Llama 4 Scout· Meta	0.5	$0.10