Inference
EN
ARC Prize Foundation

ARC-AGI-1 (public eval)

Public set of abstract reasoning puzzles.

ModelsModels with a result.116
Top resultBest result on this benchmark.99.5%Gemini 4 Argon
Top-3 spreadPoints from first to third.0.5 pts
YearYear of release.—

Result and cost

116 of 116 results

Results

116 results
#ModelResultThe score from the source.CostUS dollars to run one task.
01Gemini 4 Argonhigh effort99.5%$0.128
02Claude Fable 5.2max effort99.0%$0.709
02Claude Fable 5.1max effort99.0%$0.893
02Claude Opus 5max effort99.0%$0.42
02GPT-5.6 Solmax effort99.0%$0.37
06Gemini 4 Carbonhigh effort98.5%$0.281
06GPT-6.1 Solmax effort98.5%$0.077
08Claude Opus 5.5max effort98.3%$0.39
08GPT-5.4 Proxhigh effort98.3%$5.38
08GPT-5.6 Terramax effort98.3%$0.32
11DeepSeek V4.1 Flash 1017max effort98.0%$0.045
11Gemini 3.8 Flashhigh effort98.0%$0.13
11GPT-5.5 Proxhigh effort98.0%$3.27
14GPT-6 Astramax effort97.8%$0.33
14Claude Sonnet 5.5max effort97.8%$0.299
14GPT-6 Solmax effort97.8%$0.111
17Claude Fable 5max effort97.6%$1.41
17GPT-5.2 Proxhigh effort97.6%$7.72
19GPT-5.5xhigh effort97.5%$0.52
20Inkling 2 Smallxhigh effort97.4%$0.062
21Gemini 3.1 Pro Preview97.2%$0.373
22Claude 4.7max effort97.0%$1.77
23Claude Opus 4.6 (120K, Max)96.8%$1.30
24Gemini 3.7 Flashhigh effort96.6%$0.091
25GPT-5.4xhigh effort96.4%$0.43
26Grok 4.6xhigh effort96.3%$0.24
26Claude Opus 4.6 (120K, High)96.3%$0.873
28Gemini 3.5 Flashhigh effort96.0%$0.32
28Gemini 3.6 Flashhigh effort96.0%$0.254
30Claude Sonnet 4.6max effort95.8%$1.09
31Grok 4.20 (Reasoning)95.5%$0.24
32Grok 4.7xhigh effort95.1%$0.487
33GPT-5.2xhigh effort95.0%$0.65
34Kimi K3max effort94.9%$0.545
35Grok 4.5high effort94.8%$0.24
35DeepSeek V4 Flash 0731max effort94.8%$0.015
35Claude Opus 4.6 (120K, Medium)94.8%$0.618
38Muse Spark Openxhigh effort94.0%$0.22
39DeepSeek V4 Pro 0813max effort93.3%$0.236
40GPT-6 Lunamax effort92.5%$0.013
40GLM-5.3-Flashmax effort92.5%$0.033
42GPT-5.6 Luna 2026-07-30max effort91.3%$0.042
43GPT-5.6 Lunamax effort90.8%$0.18
44Inkling90.6%$0.216
45Claude Haiku 5.5max effort90.1%$0.038
46Claude Opus 4.6 (120K, Low)89.6%$0.411
47Inkling-Smallxhigh effort89.4%$0.084
48Gemini 3 Flash Previewhigh effort88.3%$0.135
49Qwen3.8-27Bxhigh effort87.5%$0.142
50Claude Opus 4.5 Thinking16k effort81.6%$0.52
51GLM-5.280.4%$0.175
52Beam 1.0xhigh effort79.1%$0.233
53GPT-5.1 (Thinking, High)77.1%$0.546
54GPT-5 Pro77.0%$4.03
55GPT-5.4 minixhigh effort75.1%$0.35
56Claude Sonnet 4.5 (Thinking 32K)73.8%$0.442
57Grok 4 (Thinking)73.7%$0.818
58Kimi K2.573.1%$0.14
59GPT-5.1 (Thinking, Medium)68.9%$0.256
60o4-minihigh effort68.0%$0.325
61Gemini 3.5 Flash-Litehigh effort66.4%$0.077
62GPT-5high effort65.9%$0.392
63o3high effort64.3%$0.404
64Claude Sonnet 4.5 (Thinking 16K)63.6%$0.242
65o3-prohigh effort63.3%$3.92
66Claude Haiku 4.5 (Thinking 32K)62.9%$0.214
67DeepSeek V3.261.6%$0.07
68GPT-5 minihigh effort61.5%$0.094
69MiniMax M2.559.1%$0.09
70GLM-558.6%$0.14
71Grok 4 Fast (Reasoning)58.1%$0.028
72Claude Sonnet 4 (Thinking 16K)56.8%$0.322
73Gemini 2.5 Pro (Thinking 16K)56.4%$0.396
74Gemini 2.5 Pro (Thinking 32K)55.9%$0.416
75Claude Opus 4 (Thinking 16K)54.3%$1.14
76Claude Sonnet 4.5 (Thinking 8K)53.5%$0.168
77Claude Haiku 4.5 (Thinking 16K)51.4%$0.098
78Claude Sonnet 4 (Thinking 8K)48.6%$0.194
79GPT-5.4 nanoxhigh effort47.9%$0.10
80o3-minihigh effort46.6%$0.305
81Claude Opus 4 (Thinking 8K)45.6%$0.782
82Claude Haiku 4.5 (Thinking 8K)45.0%$0.064
83Gemini 2.5 Pro (Thinking 8K)44.2%$0.261
84GPT-5.1 (Thinking, Low)44.0%$0.093
84Tiny Recursion Model (TRM)44.0%$0.395
86Claude Opus 4 (Thinking 1K)43.3%$0.424
87Nemotron 3.5 Super VLhigh effort39.8%$0.027
88Gemini 2.5 Flash (Preview) (Thinking 24K)39.1%$0.186
89Claude Sonnet 4.5 (Thinking 1K)36.6%$0.082
90Gemini 2.5 Flash (Preview)36.3%$0.033
91Gemini 2.5 Flash (Preview) (Thinking 16K)36.0%$0.185
92Claude Opus 435.5%$0.349
93Claude Sonnet 4.535.4%$0.069
94Gemini 2.5 Flash (Preview) (Thinking 8K)34.5%$0.116
95Codex Mini (Latest)33.4%$0.127
96Claude Sonnet 433.0%$0.07
97Claude Sonnet 4 (Thinking 1K)31.3%$0.081
98GPT-5 nanohigh effort29.7%$0.025
99Claude Haiku 4.5 (Thinking 1K)27.1%$0.027
100DeepSeek-R1-052827.0%$0.047
101Claude Haiku 4.526.6%$0.022
102Gemini 2.5 Flash (Preview) (Thinking 1K)18.0%$0.023
103Grok 3 Minilow effort17.6%$0.0087
104Gemini 2.5 Pro (Thinking 1K)17.5%$0.048
105Qwen3 235B 250717.0%$0.0021
106o1-mini13.2%$0.124
107GPT-5.1 Thinkingnone effort12.4%$0.028
108GPT-4.111.8%$0.032
109Magistral Medium8.9%$0.107
110Magistral Small8.6%$0.029
111Grok 3 [Beta]8.4%$0.074
112Magistral Medium (Thinking)8.0%$0.106
113GPT-4.1 mini7.2%$0.0066
114Llama 4 Maverick7.1%$0.0065
115Llama 4 Scout2.4%$0.0036
116GPT-4.1 nano1.8%$0.0018