ARC-AGI
The original ARC abstraction and reasoning corpus.
Agent Index weight
0%
shown for context only
Models tested
95
from the benchmark’s leaderboard and reports
Results
| # | Model | Score | Normalised | Harness | Cost/task | Date |
|---|---|---|---|---|---|---|
| 1 | Claude Fable 5 Supported Anthropic | 98.5% | source | — | — | |
| 2 | Claude Fable 5.2 Supported Anthropic | 98.5% | source | — | — | |
| 3 | Claude Opus 5.5 Supported Anthropic | 98.5% | source | — | — | |
| 4 | Gemini 3.8 Flash Supported Google | 98.5% | source | — | — | |
| 5 | Gemini 4 Argon Supported Google | 98.5% | source | — | — | |
| 6 | Gemini 4 Carbon Supported Google | 98.5% | source | — | — | |
| 7 | GPT-6.1 Sol Estimated OpenAI | 98.5% | source | — | — | |
| 8 | GPT-6 Astra Supported OpenAI | 98.5% | source | — | — | |
| 9 | Gemini 3.1 Pro Preview Supported Google | 98.0% | source | $0.52 | — | |
| 10 | Claude Fable 5.1 Supported Anthropic | 97.5% | source | — | — | |
| 11 | Claude Opus 5 Supported Anthropic | 97.5% | source | — | — | |
| 12 | GPT-5.6 Sol Supported OpenAI | 97.5% | source | $0.40 | — | |
| 13 | Claude Sonnet 5.5 Estimated Anthropic | 96.5% | source | — | — | |
| 14 | GPT-5.5 Pro OpenAI | 96.5% | source | $4.53 | — | |
| 15 | GPT-5.6 Terra Estimated OpenAI | 96.5% | source | $0.55 | — | |
| 16 | Inkling 2 Small Estimated thinkingmachines | 96.5% | source | — | — | |
| 17 | Gemini 3 Deep Think Google | 96.0% | source | — | — | |
| 18 | Gemini 3.7 Flash Estimated Google | 95.5% | source | — | — | |
| 19 | GPT-6 Sol Estimated OpenAI | 95.5% | source | — | — | |
| 20 | GPT-5.5 Supported OpenAI | 95.0% | source | — | — | |
| 21 | DeepSeek V4.1 Flash Estimated DeepSeek | 94.5% | source | — | — | |
| 22 | DeepSeek V4.1 Flash 1017 Estimated DeepSeek | 94.5% | source | — | — | |
| 23 | GPT-5.4 Pro OpenAI | 94.5% | source | — | — | |
| 24 | Kimi K3 Supported Moonshot AI | 94.5% | source | $0.77 | — | |
| 25 | Claude Opus 4.6 Supported Anthropic | 94.0% | source | — | — | |
| 26 | GPT-5.4 Supported OpenAI | 93.7% | source | — | — | |
| 27 | Claude Opus 4.7 Supported Anthropic | 93.5% | source | $1.41 | — | |
| 28 | Claude Opus 4.8 Supported Anthropic | 92.5% | source | $2.33 | — | |
| 29 | Gemini 3.5 Flash Supported Google | 92.5% | source | $0.43 | — | |
| 30 | Gemini 3.6 Flash Estimated Google | 91.2% | source | — | — | |
| 31 | GLM-5.3-Flash Supported Z.ai (Zhipu) | 91.0% | source | — | — | |
| 32 | DeepSeek V4 Pro 0813 DeepSeek | 90.5% | source | — | — | |
| 33 | GPT-5.2 Pro OpenAI | 90.5% | source | $11.65 | — | |
| 34 | Grok 4.7 Estimated xAI | 90.2% | source | — | — | |
| 35 | Grok 4.20 (Non-Reasoning) Estimated xAI | 89.5% | source | $0.35 | — | |
| 36 | Claude Haiku 5.5 Estimated Anthropic | 89.0% | source | — | — | |
| 37 | DeepSeek V4 Flash 0731 DeepSeek | 89.0% | source | — | — | |
| 38 | Muse Spark Open Meta | 89.0% | source | — | — | |
| 39 | GPT-5.6 Luna Estimated OpenAI | 88.0% | source | $0.32 | — | |
| 40 | Grok 4.6 Supported xAI | 87.5% | source | — | — | |
| 41 | Qwen3.8 27B Alibaba (Qwen) | 87.5% | source | — | — | |
| 42 | Grok 4.5 Estimated xAI | 87.2% | source | $0.31 | — | |
| 43 | GPT-6 Luna OpenAI | 86.7% | source | — | — | |
| 44 | Claude Sonnet 4.6 Supported Anthropic | 86.5% | source | $1.19 | — | |
| 45 | GPT-5.2 Supported OpenAI | 86.2% | source | $0.96 | — | |
| 46 | Gemini 3 Flash Preview Supported Google | 84.7% | source | $0.17 | — | |
| 47 | Inkling Small thinkingmachines | 84.0% | source | — | — | |
| 48 | Claude Opus 4.5 (latest) Supported Anthropic | 80.0% | source | $1.47 | — | |
| 49 | Inkling Estimated thinkingmachines | 79.5% | source | $0.30 | — | |
| 50 | GLM-5.2 Supported Z.ai (Zhipu) | 77.0% | source | $0.19 | — | |
| 51 | Gemini 3 Pro Preview Supported Google | 75.0% | source | $0.49 | — | |
| 52 | Beam 1.0 Reflection | 74.0% | source | — | — | |
| 53 | GPT-5.1 Supported OpenAI | 72.8% | source | — | — | |
| 54 | GPT-5 Pro OpenAI | 70.2% | source | — | — | |
| 55 | Grok 4 Supported xAI | 66.7% | source | $2.17 | — | |
| 56 | GPT-5 Supported OpenAI | 65.7% | source | $0.51 | — | |
| 57 | Kimi K2.5 Supported Moonshot AI | 65.3% | source | $0.15 | — | |
| 58 | Claude Sonnet 4.5 (latest) Supported Anthropic | 63.7% | source | $0.52 | — | |
| 59 | GPT-5.4 mini OpenAI | 63.7% | source | $0.51 | — | |
| 60 | MiniMax-M2.5 Estimated MiniMax | 63.7% | source | $0.07 | — | |
| 61 | o3 Supported OpenAI | 60.8% | source | $0.50 | — | |
| 62 | o3-pro OpenAI | 59.3% | source | $4.16 | — | |
| 63 | o4-mini Estimated OpenAI | 58.7% | source | $0.41 | — | |
| 64 | DeepSeek V3.2 Supported DeepSeek | 57.0% | source | $0.08 | — | |
| 65 | GPT-5 Mini Estimated OpenAI | 54.3% | source | $0.12 | — | |
| 66 | Gemini 3.5 Flash Lite Google | 53.5% | source | — | — | |
| 67 | GPT-5.4 nano OpenAI | 51.5% | source | $0.13 | — | |
| 68 | Grok 4 Fast xAI | 48.5% | source | $0.03 | — | |
| 69 | Claude Haiku 4.5 Estimated Anthropic | 47.7% | source | $0.26 | — | |
| 70 | GLM-5 Supported Z.ai (Zhipu) | 44.7% | source | $0.17 | — | |
| 71 | Gemini 2.5 Pro Supported Google | 41.0% | source | $0.48 | — | |
| 72 | Claude Sonnet 4 Supported Anthropic | 40.0% | source | $0.37 | — | |
| 73 | tiny-recursion-model Unknown | 40.0% | source | $1.76 | — | |
| 74 | Claude-Opus-4 Supported Anthropic | 35.7% | source | $1.25 | — | |
| 75 | o3-mini Estimated OpenAI | 34.5% | source | $0.40 | — | |
| 76 | Gemini 2.5 Flash Estimated Google | 33.3% | source | $0.21 | — | |
| 77 | o1 Estimated OpenAI | 30.7% | source | — | — | |
| 78 | Claude 3.7 Sonnet Supported Anthropic | 28.6% | source | $0.33 | — | |
| 79 | codex-mini-2025-05-16 Unknown | 27.3% | source | $0.16 | — | |
| 80 | o1-pro OpenAI | 23.3% | source | — | — | |
| 81 | DeepSeek-R1 Estimated DeepSeek | 21.2% | source | $0.05 | — | |
| 82 | GPT-5 Nano OpenAI | 20.7% | source | $0.01 | — | |
| 83 | Grok-3 mini xAI | 16.5% | source | — | — | |
| 84 | o1-mini OpenAI | 14.0% | source | — | — | |
| 85 | Qwen3 235B A22B Instruct 2507 Alibaba (Qwen) | 11.0% | source | $0.00 | — | |
| 86 | GPT-4.5 OpenAI | 10.3% | source | $0.29 | — | |
| 87 | Magistral Medium 1.1 Mistral | 6.1% | source | $0.12 | — | |
| 88 | GPT-4.1 OpenAI | 5.5% | source | $0.04 | — | |
| 89 | Grok 3 xAI | 5.5% | source | $0.09 | — | |
| 90 | Magistral Small 1.0 Mistral | 5.0% | source | $0.04 | — | |
| 91 | GPT-4o (2024-08-06) Supported OpenAI | 4.5% | source | $0.05 | — | |
| 92 | Llama 4 Maverick Meta | 4.4% | source | $0.01 | — | |
| 93 | GPT-4.1 mini OpenAI | 3.5% | source | $0.01 | — | |
| 94 | Llama 4 Scout Meta | 0.5% | source | $0.00 | — | |
| 95 | GPT-4.1 nano OpenAI | 0.0% | source | $0.00 | — |