Inference
EN
ARC Prize Foundation

ARC-AGI-2 (public eval)

Public set of abstract reasoning puzzles.

ModelsModels with a result.122
Top resultBest result on this benchmark.98.8%Claude Fable 5.2
Top-3 spreadPoints from first to third.0.8 pts
YearYear of release.—

Result and cost

122 of 122 results

Results

122 results
#ModelResultThe score from the source.CostUS dollars to run one task.
01Claude Fable 5.2max effort98.8%$2.69
02Claude Fable 5.1max effort98.3%$4.33
03Gemini 4 Carbonhigh effort97.9%$0.753
03Claude Opus 5.5max effort97.9%$1.90
05GPT-6 Astramax effort97.5%$0.97
06Gemini 4 Argonhigh effort97.1%$0.356
06GPT-6.1 Solmax effort97.1%$0.226
06Claude Opus 5max effort97.1%$2.07
09Claude Fable 5max effort96.4%$5.69
10GPT-5.6 Solmax effort93.8%$1.52
11Gemini 3.8 Flashhigh effort92.5%$0.418
12GPT-5.4 Proxhigh effort92.2%$17.60
13GPT-5.6 Terramax effort91.3%$1.18
14GPT-5.5xhigh effort90.6%$1.91
15GPT-5.5 Proxhigh effort90.4%$11.10
16Claude Sonnet 5.5xhigh effort90.1%$0.46
17GPT-6 Solmax effort90.0%$0.414
18Gemini 3.1 Pro Preview88.1%$0.979
19Gemini 3.7 Flashhigh effort85.3%$0.273
20GPT-5.4xhigh effort84.2%$1.57
20Inkling 2 Smallxhigh effort84.2%$0.269
22Claude 4.7max effort82.2%$8.21
23DeepSeek V4.1 Flash 1017max effort81.7%$0.137
24Claude Opus 4.6 (120K, High)79.0%$3.81
25Claude Opus 4.6 (120K, Max)74.9%$3.77
26Claude Opus 4.6 (120K, Medium)73.6%$3.02
27Gemini 3.5 Flashhigh effort72.1%$0.90
28Grok 4.6xhigh effort68.3%$0.755
29GLM-5.3-Flashmax effort66.8%$0.099
30Gemini 3.6 Flashhigh effort65.3%$0.658
31Claude Haiku 5.5xhigh effort64.4%$0.093
32GPT-5.6 Luna 2026-07-30max effort64.2%$0.19
33DeepSeek V4 Flash 0731max effort63.9%$0.044
34Grok 4.20 (Reasoning)63.6%$0.96
35Claude Sonnet 4.6max effort62.4%$2.89
36GPT-6 Lunamax effort61.8%$0.062
37Kimi K3max effort61.7%$1.66
38GPT-5.6 Lunamax effort60.2%$0.69
39Grok 4.7xhigh effort60.0%$2.04
40Claude Opus 4.6 (120K, Low)59.9%$2.40
41GPT-5.2xhigh effort59.8%$2.05
42DeepSeek V4 Pro 0813max effort59.4%$0.638
43Grok 4.5high effort58.2%$0.81
44Muse Spark Openxhigh effort57.2%$0.807
45GPT-5.2 Prohigh effort51.7%$16.66
46Inkling-Smallxhigh effort39.3%$0.258
47Qwen3.8-27Bxhigh effort38.8%$0.473
48Inkling37.8%$0.713
49Gemini 3 Flash Previewhigh effort34.0%$0.24
50Beam 1.0xhigh effort26.3%$0.627
51Claude Opus 4.5 Thinking16k effort24.2%$0.82
52Grok 4 (Thinking)21.1%$2.08
53GLM-5.220.8%$0.229
54GPT-5.1 (Thinking, High)18.3%$1.22
55GPT-5.4 minixhigh effort17.8%$0.81
56Claude Sonnet 4.5 (Thinking 32K)14.7%$0.84
57GPT-5 Pro13.3%$8.01
58Kimi K2.512.1%$0.30
59GPT-5high effort9.6%$0.774
60GPT-5.1 (Thinking, Medium)8.4%$0.46
61Gemini 3.5 Flash-Litehigh effort7.9%$0.147
62o4-minihigh effort7.5%$0.882
63Claude Sonnet 4.5 (Thinking 16K)6.5%$0.356
64GPT-5 minihigh effort5.8%$0.215
65GLM-55.4%$0.29
65MiniMax M2.55.4%$0.19
67Claude Haiku 4.5 (Thinking 32K)5.1%$0.393
67Gemini 2.5 Pro (Thinking 16K)5.1%$0.722
69Tiny Recursion Model (TRM)5.0%$1.80
70Claude Sonnet 4.5 (Thinking 8K)4.9%$0.241
71Grok 4 Fast (Reasoning)4.6%$0.061
71Gemini 2.5 Pro (Thinking 32K)4.6%$0.763
73Claude Sonnet 4 (Thinking 16K)4.4%$0.51
74GPT-5.4 nanoxhigh effort4.2%$0.21
74Claude Haiku 4.5 (Thinking 16K)4.2%$0.143
76DeepSeek V3.23.9%$0.13
76o3-prohigh effort3.9%$9.15
78Claude Sonnet 4.53.8%$0.131
79Claude Opus 4 (Thinking 16K)3.3%$2.02
80o3high effort2.9%$0.90
80Claude Sonnet 4 (Thinking 8K)2.9%$0.272
82Gemini 2.5 Flash (Preview) (Thinking 24K)2.6%$0.331
82Gemini 2.5 Pro (Thinking 8K)2.6%$0.449
84Claude Opus 4 (Thinking 8K)2.5%$1.20
85GPT-5.1 (Thinking, Low)2.2%$0.135
86Claude Sonnet 42.1%$0.131
86Claude Haiku 4.5 (Thinking 8K)2.1%$0.095
86Claude Sonnet 4.5 (Thinking 1K)2.1%$0.142
89Nemotron 3.5 Super VLhigh effort1.9%$0.044
89Gemini 2.5 Flash (Preview)1.9%$0.059
91Claude Opus 41.3%$0.663
91Claude Sonnet 4 (Thinking 1K)1.3%$0.147
93Gemini 2.5 Flash (Preview) (Thinking 16K)1.1%$0.324
93Gemini 2.5 Flash (Preview) (Thinking 8K)1.1%$0.205
95Qwen3 235B 25070.8%$0.0045
95Claude 3.716k effort0.8%$0.57
97Claude Opus 4 (Thinking 1K)0.4%$0.758
97Gemini 2.5 Pro (Thinking 1K)0.4%$0.088
99GPT-5 nanohigh effort0.3%$0.032
99DeepSeek-R10.3%$0.08
99DeepSeek-R1-05280.3%$0.05
102Grok 3 Minilow effort0.0%$0.013
102Claude Haiku 4.50.0%$0.043
102o3-minihigh effort0.0%$0.607
102Grok 3 [Beta]0.0%$0.14
102GPT-4.50.0%$2.07
102GPT-4.10.0%$0.069
102GPT-4.1 mini0.0%$0.014
102Llama 4 Maverick0.0%$0.013
102Magistral Medium0.0%$0.106
102o1-mini0.0%$0.184
102Magistral Small0.0%$0.05
102GPT-4o0.0%$0.08
102Llama 4 Scout0.0%$0.006
102GPT-4.1 nano0.0%$0.0036
102GPT-4o mini0.0%$0.01
102Claude Haiku 4.5 (Thinking 1K)0.0%$0.048
102Codex Mini (Latest)0.0%$0.227
102Gemini 1.5 Pro0.0%$0.04
102Gemini 2.5 Flash (Preview) (Thinking 1K)0.0%$0.035
102GPT-5.1 Thinkingnone effort0.0%$0.059
102Magistral Medium (Thinking)0.0%$0.124