Inference
ARC Prize Foundation

ARC-AGI-2 (semi-private)

Abstract reasoning puzzles held out from training data.

ModelsModels with a result.113
Top resultBest result on this test.95.0%GPT-6 Astra
Top-3 spreadPoints from first to third.4.6 pts
YearYear of release.N/A

Result and cost

020406080100$0.003$0.01$0.03$0.1$0.3$1$3$10$30$100RESULTUS DOLLARS PER TASK · LOG SCALE

Results

113 results
#ModelResultThe score from the source.CostUS dollars to run one task.
01GPT-6 Astramax effort95.0%$1.12
02GPT-5.6 Solmax effort92.5%$1.44
03Claude Opus 5max effort90.4%$2.06
04Claude Fable 5.1max effort90.0%$4.49
05Claude Fable 5max effort89.2%$5.45
06GPT-5.5xhigh effort85.0%$1.87
07Gemini 3.7 Flashhigh effort84.6%$0.249
08Gemini 3 Deep Think (2/26)84.6%$13.62
09GPT-5.5 Proxhigh effort84.2%$10.76
10GPT-5.6 Terramax effort83.9%$1.09
11GPT-5.4 Proxhigh effort83.3%$16.41
12Gemini 3.1 Pro Preview77.1%$0.962
13dots3-note Previewmax effort76.8%$0.077
14Claude 4.7max effort75.8%$7.43
15GPT-5.4xhigh effort74.0%$1.52
16GPT-5.2 (Refine.)72.9%$38.99
17Claude Opus 4.8high effort72.1%$2.74
17Gemini 3.5 Flashhigh effort72.1%$0.85
19Claude Opus 4.6 (120K, High)69.2%$3.47
20Claude Opus 4.6 (120K, Max)68.8%$3.64
21Grok 4.6xhigh effort67.1%$0.757
22Claude Opus 4.6 (120K, Medium)66.3%$2.72
23Grok 4.20 (Reasoning)65.1%$0.92
24Claude Opus 4.6 (120K, Low)64.6%$2.25
25Gemini 3.6 Flashhigh effort60.4%$0.606
26GPT-5.6 Luna 2026-07-30max effort59.6%$0.177
27GPT-5.6 Lunamax effort59.5%$0.67
28Claude Sonnet 4.6max effort58.3%$2.72
29GPT-5.2 Prohigh effort54.2%$15.72
30Gemini 3 Pro (Refine.)54.0%$30.57
31GPT-5.2xhigh effort52.9%$1.9
32Grok 4.5high effort52.6%$0.78
33Gemini 3 Deep Think (Preview) ²45.1%$77.16
34Inkling-Smallxhigh effort40.1%$0.233
35Opus 4.5 (Thinking, 64K)37.6%$2.4
36Inkling36.5%$0.642
37Gemini 3 Flash Previewhigh effort33.6%$0.231
38Gemini 3 Pro31.1%$0.811
39Grok 4 (Refine.)29.4%$30.4
40Opus 4.5 (Thinking, 16K)22.8%$0.79
41GLM-5.222.8%$0.25
42GPT-5.4 minixhigh effort18.9%$0.75
43GPT-5 Pro18.3%$7.14
44GPT-5.1 (Thinking, High)17.6%$1.17
45Grok 4 (Thinking)16.0%$2.17
46Opus 4.5 (Thinking, 8K)13.9%$0.48
47Claude Sonnet 4.5 (Thinking 32K)13.6%$0.759
48Kimi K2.511.8%$0.28
49Gemini 3.5 Flash-Litehigh effort10.3%$0.139
50GPT-5high effort9.9%$0.73
51Claude Opus 4 (Thinking 16K)8.6%$1.93
52Opus 4.5 (Thinking, None)7.8%$0.22
53Claude Sonnet 4.5 (Thinking 16K)6.9%$0.35
53Claude Sonnet 4.5 (Thinking 8K)6.9%$0.235
55o3high effort6.5%$0.834
55GPT-5.1 (Thinking, Medium)6.5%$0.421
57Tiny Recursion Model (TRM)6.3%$2.1
58o4-mini (high)high effort6.1%$0.856
59Claude Sonnet 4 (Thinking 16K)5.9%$0.486
60Claude Sonnet 4.5 (Thinking 1K)5.8%$0.142
61GPT-5.4 nanoxhigh effort5.7%$0.16
62Grok 4 Fast (Reasoning)5.3%$0.061
63GLM-54.9%$0.27
63MiniMax M2.54.9%$0.17
63Gemini 2.5 Pro (Thinking 32K)4.9%$0.757
63o3-prohigh effort4.9%$7.55
67Claude Opus 4 (Thinking 8K)4.5%$1.16
68GPT-5 minihigh effort4.4%$0.198
69DeepSeek V3.24.0%$0.12
69Claude Haiku 4.5 (Thinking 32K)4.0%$0.377
69Gemini 2.5 Pro (Thinking 16K)4.0%$0.715
72Claude Sonnet 4.53.8%$0.13
73o3-minihigh effort3.0%$0.547
74Gemini 2.5 Pro (Thinking 8K)2.9%$0.444
75Claude Haiku 4.5 (Thinking 16K)2.8%$0.139
76GPT-5 nanohigh effort2.6%$0.03
77Gemini 2.5 Flash (Preview) (Thinking 24K)2.5%$0.319
78Gemini 2.5 Flash (Preview) (Thinking 1K)2.2%$0.03
79Claude Sonnet 4 (Thinking 8K)2.1%$0.265
79Gemini 2.5 Flash (Preview) (Thinking 8K)2.1%$0.199
81Gemini 2.5 Flash (Preview) (Thinking 16K)2.0%$0.317
82GPT-5.1 (Thinking, Low)1.9%$0.129
83Gemini 2.5 Flash (Preview)1.7%$0.057
84Claude Haiku 4.5 (Thinking 8K)1.7%$0.091
85DeepSeek-R11.3%$0.08
85Gemini 2.0 Flash1.3%<$0.01
87Claude Opus 41.3%$0.639
87Claude Sonnet 41.3%$0.127
87Codex Mini (Latest)1.3%$0.23
90Claude Haiku 4.51.3%$0.043
90Claude Haiku 4.5 (Thinking 1K)1.3%$0.047
90Qwen3-235b-a22b Instruct (25/07)1.3%<$0.01
93Deepseek R1 (05/28)1.1%$0.053
94Claude Sonnet 4 (Thinking 1K)0.9%$0.143
95O1 Mini0.8%$0.191
96Gemini 1.5 Pro0.8%$0.04
96GPT-4.50.8%$2.1
98Claude 3.716k effort0.7%$0.51
99GPT-4.10.4%$0.069
99GPT-5.1 (Thinking, None)0.4%$0.058
99Grok 3 Minilow effort0.4%$0.013
102Grok 3 [Beta]0.0%$0.142
102Magistral Medium0.0%$0.108
102Magistral Small0.0%$0.049
102GPT-4.1 mini0.0%$0.014
102Llama 4 Maverick0.0%$0.012
102GPT-4o0.0%$0.08
102Llama 4 Scout0.0%<$0.01
102GPT-4.1 nano0.0%<$0.01
102GPT-4o mini0.0%$0.01
102Claude Opus 4 (Thinking 1K)0.0%$0.75
102Gemini 2.5 Pro (Thinking 1K)0.0%$0.089
102Magistral Medium (Thinking)0.0%$0.123