Benchmark · ReasoningSettled

ARC-AGI

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

Updated 2026-11-10
Models tested
90
Top score
98.5
Claude Fable 5
Median
78.5
min 0.1
Top-5 spread
σ 0.0
Settled

Best score over time · one chart, every benchmark

ARC-AGI80 MODELS · FRONTIER RUNNING MAX0255075100SCORE ↑Nov 24May 25Nov 25May 26Nov 26RELEASE DATE →benchgecko.ai/benchmark/arc-agi · frontier
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 · 14 totalClick to open model page

90 models tested · sorted by score

#ModelScore
1Anthropic logoClaude Fable 598.5
2OpenAI logoGPT-6 Astra98.5
3Google DeepMind logoGemini 4 Carbon98.5
4Anthropic logoClaude Opus 5.598.5
5OpenAI logoGPT-6.1 Sol98.5
6Google DeepMind logoGemini 3.8 Flash98.5
7Google DeepMind logoGemini 4 Argon98.5
8Anthropic logoClaude Fable 5.298.5
9Google DeepMind logoGemini 3.1 Pro Preview98.0
10Anthropic logoClaude Fable 5.197.5
11Anthropic logoClaude Opus 597.5
12OpenAI logoGPT-5.6 Sol97.5
13OpenAI logoGPT-5.6 Terra96.5
14OpenAI logoGPT-5.5 Pro96.5
15Anthropic logoClaude Sonnet 5.596.5
16Google DeepMind logoGemini 3.7 Flash95.5
17OpenAI logoGPT-6 Sol95.5
18OpenAI logoGPT-5.595.0
19OpenAI logoGPT-5.4 Pro94.5
20moonshotai logoKimi K394.5
21DeepSeek logoDeepSeek V4.1 Flash94.5
22DeepSeek logoDeepSeek V4.1 Flash 101794.5
23Anthropic logoClaude Opus 4.694.0
24OpenAI logoGPT-5.493.7
25Anthropic logoClaude Opus 4.793.5
26Anthropic logoClaude Opus 4.892.5
27Google DeepMind logoGemini 3.5 Flash92.5
28Google DeepMind logoGemini 3.6 Flash91.2
29z-ai logoGLM 5.3 Flash91.0
30OpenAI logoGPT-5.2 Pro90.5
31DeepSeek logoDeepSeek V4 Pro 081390.5
32xAI logoGrok 4.790.2
33xAI logoGrok 4.2089.5
34DeepSeek logoDeepSeek V4 Flash 073189.0
35Meta logoMuse Spark Open89.0
36Anthropic logoClaude Haiku 5.589.0
37OpenAI logoGPT-5.6 Luna88.0
38Alibaba Qwen logoQwen3.8 27B87.5
39xAI logoGrok 4.687.5
40xAI logoGrok 4.587.2
41OpenAI logoGPT-6 Luna86.7
42Anthropic logoClaude Sonnet 4.686.5
43OpenAI logoGPT-5.286.2
44Google DeepMind logoGemini 3 Flash Preview84.7
45Anthropic logoClaude Opus 4.580.0
46z-ai logoGLM 5.277.0
47Google DeepMind logoGemini 3 Pro75.0
48OpenAI logoGPT-5.172.8
49OpenAI logoGPT-5 Pro70.2
50xAI logoGrok 466.7
51OpenAI logoGPT-565.7
52moonshotai logoKimi K2.565.3
53Anthropic logoClaude Sonnet 4.563.7
54minimax logoMiniMax M2.563.7
55OpenAI logoGPT-5.4 Mini63.7
56OpenAI logoo360.8
57OpenAI logoo3 Pro59.3
58OpenAI logoo4 Mini58.7
59OpenAI logoo4 Mini (batch)58.7
60DeepSeek logoDeepSeek V3.257.0
61OpenAI logoGPT-5 Mini54.3
62Google DeepMind logoGemini 3.5 Flash Lite53.5
63OpenAI logoGPT-5.4 Nano51.5
64xAI logoGrok 4 Fast48.5
65Anthropic logoClaude Haiku 4.547.7
66z-ai logoGLM 544.7
67Google DeepMind logoGemini 2.5 Pro41.0
68Anthropic logoClaude Sonnet 440.0
69Anthropic logoClaude Opus 435.7
70OpenAI logoo3 Mini (batch)34.5
71OpenAI logoo3 Mini34.5
72Google DeepMind logoGemini 2.5 Flash32.3
73OpenAI logoo130.7
74Anthropic logoClaude 3.7 Sonnet28.6
75DeepSeek logoR1 052821.2
76OpenAI logoGPT-5 Nano20.7
77OpenAI logoo1-preview18.0
78xAI logoGrok-3 mini16.5
79DeepSeek logoR115.8
80OpenAI logoo1-mini14.0
81Alibaba Qwen logoQwen3 235B A22B Instruct 250711.0
82OpenAI logoGPT-4.510.3
83xAI logoGrok 35.5
84OpenAI logoGPT-4.15.5
85Magistral Small 1.15.0
86OpenAI logoGPT-4o (2024-11-20)4.5
87Meta logoLlama 4 Maverick4.4
88OpenAI logoGPT-4.1 Mini3.5
89Meta logoLlama 4 Scout0.5
90OpenAI logoGPT-4.1 Nano0.1

Same category · related evaluations