Skip to main content

AI Rankings

ARC-AGI-2 Leaderboard: AI Models Ranked by Score

This page ranks AI models by their ARC-AGI-2 score, using Epoch AI's open benchmark data.

Data updated

What ARC-AGI-2 measures

Visual puzzles where the AI must work out a new rule from a few examples and apply it. They test picking up a new idea quickly, not recalling facts.

Scores are percentages and higher is better. Results come from Epoch AI's open benchmark data; we do not run the test ourselves.

Top 20 at a glance

Tap a model to see its full page: scores, price and how it compares.
Top 20 models on ARC-AGI-2 scoreRanked list of 20 items. 1. GPT-6 Astra, 95.0%; 2. GPT-6.1 Sol, 94.2%; 3. Claude Opus 5.5, 93.3%.
  1. GPT-6 Astra95.0%Released 3 Sep 2026
  2. GPT-6.1 Sol94.2%Released 29 Sep 2026
  3. Claude Opus 5.593.3%Released 22 Sep 2026
  4. GPT-5.6 Sol92.5%Released 9 Jul 2026
  5. Gemini 4 Argon91.3%Released 30 Sep 2026
  6. Claude Opus 590.4%Released 24 Jul 2026
  7. Claude Fable 5.190.0%Released 1 Sep 2026
  8. GPT-6 Sol89.6%Released 22 Sep 2026
  9. Claude Fable 589.2%Released 9 Jun 2026
  10. Gemini 3.8 Flash89.2%Released 2 Sep 2026
  11. GPT-5.585.0%Released 23 Apr 2026
  12. Gemini 3.7 Flash84.6%Released 13 Aug 2026
  13. GPT-5.5 Pro84.6%Released 23 Apr 2026
  14. GPT-5.6 Terra83.9%Released 9 Jul 2026
  15. GPT-5.4 Pro83.3%Released 5 Mar 2026
  16. Gemini 3.1 Pro77.1%Released 19 Feb 2026
  17. Claude Opus 4.775.8%Released 16 Apr 2026
  18. GPT-5.474.0%Released 5 Mar 2026
  19. DeepSeek V4.1 Flash72.9%Open weightsReleased 9 Sep 2026
  20. Claude Opus 4.872.1%Released 28 May 2026

Full ranking

All 86 models.Download the full table (CSV)Data credit: Epoch AI, LiteLLM model price list, used under open licences (details at the foot of the page).
ARC-AGI-2 Leaderboard: AI Models Ranked by Score: 86 of 86 models
Rank and modelARC-AGI-2 scoreEpoch indexPrice per 1M tokens (in / out)Released
1GPT-6 AstraOpenAI
95.0%168.8$10.00 / $50.002026-09-03
2GPT-6.1 SolOpenAI
94.2%168.4$2.00 / $10.002026-09-29
3Claude Opus 5.5Anthropic
93.3%168.8$4.00 / $20.002026-09-22
4GPT-5.6 SolOpenAI
92.5%162.8$4.00 / $20.002026-07-09
5Gemini 4 ArgonGoogle
91.3%166.4$2.00 / $10.002026-09-30
6Claude Opus 5Anthropic
90.4%163.9$5.00 / $25.002026-07-24
7Claude Fable 5.1Anthropic
90.0%166.5$10.00 / $50.002026-09-01
8GPT-6 SolOpenAI
89.6%163.7$2.00 / $10.002026-09-22
9Claude Fable 5Anthropic
89.2%163.4$10.00 / $50.002026-06-09
9Gemini 3.8 FlashGoogle
89.2%157.4$0.75 / $3.752026-09-02
11GPT-5.5OpenAI
85.0%159.7$5.00 / $30.002026-04-23
12Gemini 3.7 FlashGoogle
84.6%158.0$0.75 / $3.752026-08-13
12GPT-5.5 ProOpenAI
84.6%162.9$30.00 / $1802026-04-23
14GPT-5.6 TerraOpenAI
83.9%160.1$2.00 / $12.002026-07-09
15GPT-5.4 ProOpenAI
83.3%159.6$30.00 / $1802026-03-05
16Gemini 3.1 ProGoogle
77.1%155.3$2.00 / $12.002026-02-19
17Claude Opus 4.7Anthropic
75.8%156.8$5.00 / $25.002026-04-16
18GPT-5.4OpenAI
74.0%157.3$2.50 / $15.002026-03-05
19DeepSeek V4.1 FlashDeepSeek · Open weights
72.9%155.9$0.30 / $1.202026-09-09
20Claude Opus 4.8Anthropic
72.1%159.0$5.00 / $25.002026-05-28
20Gemini 3.5 FlashGoogle
72.1%154.9$1.50 / $9.002026-05-19
22Claude Opus 4.6Anthropic
69.2%155.7$5.00 / $25.002026-02-05
23Grok 4.6xAI
67.1%156.6$2.00 / $6.002026-08-12
24GLM-5.3-FlashZ.ai · Open weights
65.8%153.1$0.15 / $0.502026-08-20
25Grok 4.20xAI
65.1%152.1$1.25 / $2.502026-02-17
26DeepSeek V4 Flash 0731DeepSeek · Open weights
61.4%154.8$0.19 / $0.382026-07-31
26Grok 4.7xAI
61.4%154.4$2.00 / $6.002026-09-21
28Claude Haiku 5.5Anthropic
61.3%159.3$0.10 / $0.502026-10-07
28DeepSeek V4 Pro 0813DeepSeek · Open weights
61.3%155.6$1.32 / $3.962026-08-13
30Claude Sonnet 4.6Anthropic
60.4%152.8$3.00 / $15.002026-02-17
30Gemini 3.6 FlashGoogle
60.4%154.5$0.75 / $3.752026-07-21
30Kimi K3Moonshot AI · Open weights
60.4%157.9$3.00 / $15.002026-07-16
33GPT-5.6 LunaOpenAI
59.5%156.5$0.20 / $1.202026-07-09
34GPT-6 LunaOpenAI
59.3%156.8$0.10 / $0.502026-09-22
35GPT-5.2 ProOpenAI
54.2%156.3$21.00 / $1682025-12-11
36GPT-5.2OpenAI
52.9%153.7$1.75 / $14.002025-12-11
37Grok 4.5xAI
52.6%154.1$2.00 / $6.002026-07-08
38Qwen 3.8 27BAlibaba · Open weights
42.4%151.7$0.45 / $3.002026-08-14
39Inkling-SmallThinking Machines · Open weights
40.1%150.3$0.47 / $1.202026-07-15
40Claude Opus 4.5Anthropic
37.6%150.1$5.00 / $25.002025-11-24
41InklingThinking Machines · Open weights
36.5%148.4$1.00 / $4.052026-07-15
42Gemini 3 FlashGoogle
33.6%151.9$0.50 / $3.002025-12-17
43Gemini 3 ProGoogle
31.1%153.1$2.00 / $12.002025-11-18
44GLM-5.2Z.ai · Open weights
22.8%151.9$1.40 / $4.402026-06-16
45GPT-5.4 MiniOpenAI
18.9%148.7$0.75 / $4.502026-03-17
46GPT-5 ProOpenAI
18.3%150.6$15.00 / $1202025-10-07
47GPT-5.1OpenAI
17.6%149.8$1.25 / $10.002025-11-13
48Grok 4xAI
16.0%146.2$3.00 / $15.002025-07-09
49Claude Sonnet 4.5Anthropic
13.6%146.7$3.00 / $15.002025-09-29
50Kimi K2.5Moonshot AI · Open weights
11.8%148.1$0.60 / $3.002026-01-27
51Gemini 3.5 Flash-LiteGoogle
10.3%144.9$0.30 / $2.502026-07-21
52GPT-5OpenAI
9.9%150.0$1.25 / $10.002025-08-07
53Claude Opus 4Anthropic
8.6%142.5$15.00 / $75.002025-05-22
54o3OpenAI
6.5%146.7$2.00 / $8.002025-04-16
55o4-miniOpenAI
6.1%145.5$1.10 / $4.402025-04-16
56Claude Sonnet 4Anthropic
5.9%141.6$3.00 / $15.002025-05-22
57GPT-5.4 NanoOpenAI
5.7%146.0$0.20 / $1.252026-03-17
58Grok 4 FastxAI
5.3%144.1Not measured2025-09-19
59Gemini 2.5 Pro (Jun 2025)Google
4.9%145.1$1.25 / $10.002025-06-05
59GLM-5Z.ai · Open weights
4.9%145.8$1.00 / $3.202026-02-11
59MiniMax-M2.5MiniMax · Open weights
4.9%146.6$0.30 / $1.202026-02-12
59o3-proOpenAI
4.9%147.3$20.00 / $80.002025-06-10
63GPT-5 miniOpenAI
4.4%145.3$0.25 / $2.002025-08-07
64Claude Haiku 4.5Anthropic
4.0%142.2$1.00 / $5.002025-10-15
64DeepSeek-V3.2DeepSeek · Open weights
4.0%146.1$0.28 / $0.402025-12-01
66o3-miniOpenAI
3.0%140.2$1.10 / $4.402025-01-31
67GPT-5 nanoOpenAI
2.6%139.2$0.05 / $0.402025-08-07
68Gemini 2.5 Flash (May 2025)Google
2.5%141.4Not measured2025-05-20
69DeepSeek-R1DeepSeek · Open weights
1.3%138.8$0.55 / $2.192025-01-20
69Gemini 2.0 Flash (Feb 2025)Google
1.3%134.6$0.15 / $0.602025-02-05
71Qwen3-235B-A22B-Instruct (Jul 2025)Alibaba · Open weights
1.3%138.8$0.22 / $0.882025-07-25
72DeepSeek-R1 (May 2025)DeepSeek · Open weights
1.1%141.1$0.75 / $2.452025-05-28
73Claude 3.7 SonnetAnthropic
0.9%141.0$3.00 / $15.002025-02-24
74o1-miniOpenAI
0.8%135.7$1.10 / $4.402024-09-12
75Gemini 1.5 Pro (Sept 2024)Google
0.8%131.7Not measured2024-09-24
75GPT-4.5OpenAI
0.8%136.7$75.00 / $1502025-02-27
77GPT-4.1OpenAI
0.4%136.7$2.00 / $8.002025-04-14
77Grok-3 minixAI
0.4%140.2$0.28 / $0.892025-06-24
79GPT-4.1 miniOpenAI
0.0%134.9$0.40 / $1.602025-04-14
79GPT-4.1 nanoOpenAI
0.0%129.5$0.10 / $0.402025-04-14
79GPT-4o (Nov 2024)OpenAI
0.0%128.7$2.50 / $10.002024-11-20
79GPT-4o miniOpenAI
0.0%126.5$0.15 / $0.602024-07-18
79Grok 3xAI
0.0%138.2$3.00 / $15.002025-04-09
79Llama 4 MaverickMeta · Open weights
0.0%132.2$0.22 / $0.792025-04-06
79Llama 4 ScoutMeta · Open weights
0.0%129.6$0.10 / $0.302025-04-05
79Magistral Small 1.0Mistral AI · Open weights
0.0%133.1Not measured2025-06-10

How to read this leaderboard

Scores are percentages and higher is better. For each model we show the best result Epoch lists across its effort settings, the same rule for every model. A gap of a point or two is usually within the normal noise of a test, so treat close scores as a tie. Newer models can be missing when nobody has run the test on them yet, so check the release dates. A benchmark measures one kind of skill: use it together with the overall leaderboard rather than on its own.

Questions people ask

What does ARC-AGI-2 measure?

Visual puzzles where the AI must work out a new rule from a few examples and apply it. They test picking up a new idea quickly, not recalling facts.

Who ran these tests?

The results come from Epoch AI's open benchmark data. It combines runs by Epoch with results published by other groups, and Epoch's own pages say which is which. We do not run tests ourselves.

Does a high score mean the model is best for me?

Not by itself. ARC-AGI-2 checks one kind of ability. A model can lead here and still be slower, costlier or a poor fit for your work. Compare it on the overall leaderboard and try it on your own tasks.

Why is a model I know missing?

A model appears only when an open source has published results for it. New models can take days or weeks to be measured, and models with no published score are left out rather than guessed.

Where does the data come from?

Scores come from Epoch AI's open benchmark data (licensed CC BY 4.0). We match models across sources, rank them and show them in charts, without changing the scores. Epoch AI does not endorse this site. Prices and context windows come from the open-source LiteLLM price list (MIT licence), which records each model maker's own price page. Prices change often, so confirm on the maker's site before you buy.

How often is this leaderboard updated?

We check our open data sources every day and show the date of the data at the top of the page. A model only moves when its source publishes new results, so a number can stay the same for weeks.

More leaderboards

Different questions need different tests. Pick the one that matches your work.

Where this data comes from

  • Epoch AI: Benchmark data: Epoch AI, 'Capabilities & benchmarking', epoch.ai/benchmarks (CC BY 4.0). We match models across sources, rank them and show them in charts. Scores are not changed. Epoch AI does not endorse this site.
  • LiteLLM model price list: Prices and context windows: LiteLLM model price list (github.com/BerriAI/litellm, MIT licence). We use the model maker's own price entry when there is one. Prices change often, so always confirm on the maker's site before you buy.

Newest data on this page: 26 Oct 2026. Prices are list prices and change often, so confirm on the maker's site before you buy. Sponsors never change a score or an order. How we calculate this · Report a mistake

We use optional cookies for analytics and referral tracking. Learn more about our privacy policy