Noometry

Reasoning benchmark

ARC-AGI-1 leaderboard

As of October 2026, Claude Fable 5 has the highest published ARC-AGI-1 score on Noometry at 98.5%, out of 83 models with results.

Last verified

About ARC-AGI-1

Abstract grid puzzles: infer a transformation rule from a few examples and apply it to a new grid.

Category
Reasoning
Introduced
2019
Format
Grid puzzles
Unit
Percent (random guessing ≈ 0%)
Official site
arcprize.org

Top 15 models

  1. Claude Fable 5 98.5%
  2. Claude Opus 5.5 98.5%
  3. Gemini 3.8 Flash 98.5%
  4. GPT-6.1 Sol 98.5%
  5. GPT-6 Astra 98.5%
  6. Gemini 3.1 Pro Preview 98%
  7. Claude Fable 5.1 97.5%
  8. Claude Opus 5 97.5%
  9. GPT-5.6 Sol 97.5%
  10. GPT-5.5 Pro 96.5%
  11. GPT-5.6 Terra 96.5%
  12. Gemini 3 Deep Think 96%
  13. Gemini 3.7 Flash 95.5%
  14. GPT-6 Sol 95.5%
  15. GPT-5.5 95%

Sponsored placements are available on pages like this one. Advertise on Noometry

All results

#ModelProviderScoreSettingSourceDate
1Claude Fable 5 Anthropic98.5%maxEpoch AI
2Claude Opus 5.5 Anthropic98.5%highEpoch AI
3Gemini 3.8 Flash Google98.5%highEpoch AI
4GPT-6.1 Sol OpenAI98.5%highEpoch AI
5GPT-6 Astra OpenAI98.5%highEpoch AI
6Gemini 3.1 Pro Preview Google98%Epoch AI
7Claude Fable 5.1 Anthropic97.5%maxEpoch AI
8Claude Opus 5 Anthropic97.5%highEpoch AI
9GPT-5.6 Sol OpenAI97.5%xhighEpoch AI
10GPT-5.5 Pro OpenAI96.5%highEpoch AI
11GPT-5.6 Terra OpenAI96.5%maxEpoch AI
12Gemini 3 Deep Think Google96%Epoch AI
13Gemini 3.7 Flash Google95.5%highEpoch AI
14GPT-6 Sol OpenAI95.5%maxEpoch AI
15GPT-5.5 OpenAI95%xhighEpoch AI
16GPT-5.4 Pro OpenAI94.5%xhighEpoch AI
17Kimi K3 Moonshot AI94.5%maxEpoch AI
18Claude Opus 4.6 Anthropic94%120KEpoch AI
19GPT-5.4 OpenAI93.7%xhighEpoch AI
20Claude Opus 4.7 Anthropic93.5%highEpoch AI
21Claude Opus 4.8 Anthropic92.5%maxEpoch AI
22Gemini 3.5 Flash Google92.5%highEpoch AI
23Gemini 3.6 Flash Google91.2%highEpoch AI
24GLM-5.3-Flash Z.ai (Zhipu)91%maxEpoch AI
25DeepSeek V4 Pro DeepSeek90.5%lowEpoch AI
26GPT-5.2 Pro OpenAI90.5%xhighEpoch AI
27Grok 4.20 (Non-Reasoning) xAI89.5%Epoch AI
28DeepSeek V4 Flash DeepSeek89%maxEpoch AI
29GPT-5.6 Luna OpenAI88%maxEpoch AI
30Grok 4.6 xAI87.5%mediumEpoch AI
31Qwen3.8 27B Alibaba (Qwen)87.5%xhighEpoch AI
32Grok 4.5 xAI87.2%mediumEpoch AI
33GPT-6 Luna OpenAI86.7%maxEpoch AI
34Claude Sonnet 4.6 Anthropic86.5%highEpoch AI
35GPT-5.2 OpenAI86.2%xhighEpoch AI
36Gemini 3 Flash Preview Google84.7%Epoch AI
37Inkling-Small Thinking Machines Lab84%xhighEpoch AI
38Claude Opus 4.5 Anthropic80%64KEpoch AI
39Inkling Thinking Machines Lab79.5%Epoch AI
40GLM-5.2 Z.ai (Zhipu)77%Epoch AI
41Gemini 3 Pro Google75%Epoch AI
42GPT-5.1 OpenAI72.8%highEpoch AI
43GPT-5 Pro OpenAI70.2%highEpoch AI
44Grok 4 xAI66.7%Epoch AI
45GPT-5 OpenAI65.7%highEpoch AI
46Kimi K2.5 Moonshot AI65.3%Epoch AI
47Claude Sonnet 4.5 Anthropic63.7%32KEpoch AI
48GPT-5.4 mini OpenAI63.7%xhighEpoch AI
49MiniMax-M2.5 MiniMax63.7%Epoch AI
50o3 OpenAI60.8%highEpoch AI
51o3-pro OpenAI59.3%highEpoch AI
52o4-mini OpenAI58.7%highEpoch AI
53DeepSeek-V3.2-Exp DeepSeek57%Epoch AI
54GPT-5 Mini OpenAI54.3%highEpoch AI
55Gemini 3.5 Flash Lite Google53.5%highEpoch AI
56GPT-5.4 nano OpenAI51.5%xhighEpoch AI
57Grok 4 Fast xAI48.5%Epoch AI
58Claude Haiku 4.5 Anthropic47.7%32KEpoch AI
59GLM-5 Z.ai (Zhipu)44.7%Epoch AI
60Gemini 2.5 Pro Google41%16KEpoch AI
61Claude Sonnet 4 Anthropic40%16KEpoch AI
62Claude Opus 4 Anthropic35.7%16KEpoch AI
63o3-mini OpenAI34.5%highEpoch AI
64Gemini 2.5 Flash Google33.3%Epoch AI
65o1 OpenAI30.7%mediumEpoch AI
66Claude 3.7 Sonnet Anthropic28.6%16KEpoch AI
67o1-pro OpenAI23.3%lowEpoch AI
68DeepSeek-R1 DeepSeek21.2%Epoch AI
69GPT-5 Nano OpenAI20.7%mediumEpoch AI
70Grok-3 mini xAI16.5%lowEpoch AI
71Grok-3 mini xAI16.5%lowEpoch AI
72o1-mini OpenAI14%Epoch AI
73Qwen3 235B-A22B Alibaba (Qwen)11%Epoch AI
74GPT-4.5 OpenAI10.3%Epoch AI
75Magistral Medium Mistral AI6.1%Epoch AI
76GPT-4.1 OpenAI5.5%Epoch AI
77Grok 3 xAI5.5%Epoch AI
78Magistral Small Mistral AI5%Epoch AI
79GPT-4o OpenAI4.5%Epoch AI
80Llama 4 Maverick Meta4.4%Epoch AI
81GPT-4.1 mini OpenAI3.5%Epoch AI
82Llama 4 Scout Meta0.5%Epoch AI
83GPT-4.1 nano OpenAI0%Epoch AI

Compare the leaders

Other reasoning benchmarks

Frequently asked questions

What does ARC-AGI-1 measure?

Abstract grid puzzles: infer a transformation rule from a few examples and apply it to a new grid.

Which model has the highest ARC-AGI-1 score?

As of October 2026, Claude Fable 5 has the highest published ARC-AGI-1 score on Noometry at 98.5%, out of 83 models with results.

What is the best open-weight model on ARC-AGI-1?

Kimi K3 has the highest ARC-AGI-1 accuracy among open-weight models at 94.5%, ranking 17 of 83 overall.