ARC-AGI-2 — raws fold in 'arc_agi_2'/'ArcAGI V2' aliases; distinct version from v1/v3
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | GPT-5.6 Sol | OpenAI | 92.5 | 2 | 2026-07-24 |
| 2 | Opus 5 | Anthropic | 90.4 | 3 | 2026-07-25 |
| 3 | Claude Fable 5 (Extra High) | Anthropic | 88.3 | 1 | 2026-08-06 |
| 4 | GPT-5.5 | OpenAI | 85 | 4 | 2026-08-23 |
| 5 | Gemini 3 Deep Think | 84.6 | 2 | 2026-05-10 | |
| 6 | GPT-5.5 Pro | OpenAI | 84.2 | 1 | 2026-05-10 |
| 7 | GPT-5.6 Terra | OpenAI | 83.9 | 1 | 2026-07-10 |
| 8 | GPT-5.4 Pro | OpenAI | 83.3 | 3 | 2026-06-21 |
| 9 | Gemini 3.1 Pro | 77.1 | 7 | 2026-08-24 | |
| 10 | Claude Fable 5 | Anthropic | 76.8 | 1 | 2026-08-06 |
| 11 | Claude Opus 4.7 | Anthropic | 75.8 | 8 | 2026-07-24 |
| 12 | GPT-5.4 | OpenAI | 74 | 6 | 2026-08-23 |
| 13 | GPT-5.2 (Refine.) | OpenAI | 72.9 | 1 | 2026-05-10 |
| 14 | Gemini 3.5 Flash | 72.1 | 5 | 2026-08-24 | |
| 15 | Claude Opus 4.8 | Anthropic | 72.1 | 2 | 2026-07-24 |
| 16 | Claude Opus 4.6 | Anthropic | 68.8 | 5 | 2026-08-23 |
| 17 | Grok 4.20 | SpaceXAI | 65.1 | 1 | 2026-05-10 |
| 18 | Claude Sonnet 4.6 | Anthropic | 60.4 | 6 | 2026-08-23 |
| 19 | GPT-5.6 Luna | OpenAI | 59.6 | 7 | 2026-08-11 |
| 20 | GPT-5.2 Pro | OpenAI | 54.2 | 2 | 2026-08-23 |
| 21 | Gemini 3 Pro | 54 | 3 | 2026-08-23 | |
| 22 | Gemini 3.7 Flash | 52.9 | 1 | 2026-08-21 | |
| 23 | GPT-5.2 | OpenAI | 52.9 | 6 | 2026-08-23 |
| 24 | DeepSeek-V4-Flash | DeepSeek | 46 | 2 | 2026-08-10 |
| 25 | Muse Spark | Meta | 42.5 | 1 | 2026-08-23 |
| 26 | Inkling-Small | Thinking Machines | 40.1 | 4 | 2026-08-24 |
| 27 | Claude Opus 4.5 | Anthropic | 37.6 | 4 | 2026-08-23 |
| 28 | Inkling | Thinking Machines | 36.5 | 1 | 2026-07-18 |
| 29 | Gemini 3 Flash Preview | 33.6 | 7 | 2026-08-23 | |
| 30 | Grok4.5 | SpaceXAI | 33.1 | 1 | 2026-07-24 |
| 31 | Grok 4 (Refine.) | SpaceXAI | 29.4 | 1 | 2026-05-10 |
| 32 | Grok 4.6 | SpaceXAI | 27.6 | 1 | 2026-08-14 |
| 33 | GLM-5.2 Full Open Source | Z.ai | 22.8 | 2 | 2026-07-13 |
| 34 | GPT-5 Pro | OpenAI | 18.3 | 1 | 2026-05-10 |
| 35 | GPT-5.1 | OpenAI | 17.6 | 4 | 2026-06-20 |
| 36 | Grok 4 | SpaceXAI | 16 | 2 | 2026-08-23 |
| 37 | Claude Sonnet 4 | Anthropic | 13.6 | 2 | 2026-05-10 |
| 38 | Claude Sonnet 4.5 | Anthropic | 13.6 | 2 | 2026-05-10 |
| 39 | GPT-5.4 mini | OpenAI | 13.2 | 2 | 2026-05-15 |
| 40 | Kimi K3 | Moonshot | 12.4 | 1 | 2026-08-01 |
| 41 | Kimi K2.5 | Moonshot | 11.8 | 1 | 2026-05-10 |
| 42 | Claude 4 Opus | Anthropic | 8.6 | 3 | 2026-08-23 |
| 43 | GPT-5 | OpenAI | 7.5 | 4 | 2026-05-10 |
| 44 | o3 | OpenAI | 6.5 | 2 | 2026-08-23 |
| 45 | Tiny Recursion Model (TRM) | Bespoke | 6.3 | 1 | 2026-05-10 |
| 46 | o4-mini | OpenAI | 6.1 | 3 | 2026-05-10 |
| 47 | Grok 4 Fast | SpaceXAI | 5.3 | 1 | 2026-05-15 |
| 48 | Gemini 2.5 Pro | 4.9 | 3 | 2026-08-23 | |
| 49 | GLM-5 | Z.ai | 4.9 | 1 | 2026-05-10 |
| 50 | MiniMax M2.5 | MiniMax | 4.9 | 1 | 2026-05-10 |
| 51 | o3-pro | OpenAI | 4.9 | 1 | 2026-05-10 |
| 52 | GPT-5 mini | OpenAI | 4 | 4 | 2026-05-15 |
| 53 | Claude Haiku 4.5 | Anthropic | 4 | 1 | 2026-05-10 |
| 54 | DeepSeek-V3.2 | DeepSeek | 4 | 1 | 2026-05-10 |
| 55 | o3-mini | OpenAI | 3 | 2 | 2026-05-10 |
| 56 | Gemini 3.6 Flash | 2.6 | 1 | 2026-08-07 | |
| 57 | Gemini 2.5 Flash | 2.5 | 2 | 2026-05-10 | |
| 58 | Hierarchical Reasoning Model (HRM) | Bespoke | 2 | 1 | 2026-05-10 |
| 59 | GPT-5.4 nano | OpenAI | 1.9 | 2 | 2026-05-10 |
| 60 | Seed Oss 36B Instruct | ByteDance | 1.4 | 1 | 2026-06-15 |
| 61 | DeepSeek-R1 | DeepSeek | 1.3 | 2 | 2026-05-10 |
| 62 | Gemini 2.0 Flash (Non-Reasoning) | 1.3 | 1 | 2026-05-10 | |
| 63 | Gemini 2.0 Flash (Reasoning) | 1.3 | 1 | 2026-06-24 | |
| 64 | Codex Mini | OpenAI | 1.3 | 1 | 2026-05-10 |
| 65 | Qwen3 235B A22B Instruct 2507 | Alibaba | 1.3 | 2 | 2026-06-19 |
| 66 | Claude 3.7 | Anthropic | 0.9 | 1 | 2026-07-14 |
| 67 | GPT-5 nano | OpenAI | 0.9 | 3 | 2026-05-10 |
| 68 | Qwen3 30B A3B | Alibaba | 0.9 | 1 | 2026-06-15 |
| 69 | Gemini 3.5 Flash Lite | 0.8 | 1 | 2026-08-07 | |
| 70 | OpenAI o1-mini | OpenAI | 0.8 | 1 | 2026-05-10 |
| 71 | GPT-4.5 Preview | OpenAI | 0.8 | 1 | 2026-05-10 |
| 72 | Gemini 1.5 Pro | 0.8 | 1 | 2026-05-15 | |
| 73 | GPT-4.1 | OpenAI | 0.4 | 1 | 2026-05-10 |
| 74 | Grok 3 Mini Reasoning | SpaceXAI | 0.4 | 1 | 2026-05-10 |
| 75 | GPT-4o mini | OpenAI | 0 | 1 | 2026-05-10 |
| 76 | Magistral Small | Mistral | 0 | 1 | 2026-05-10 |
| 77 | GPT-4.1 mini | OpenAI | 0 | 2 | 2026-06-19 |
| 78 | Llama 4 Scout | Meta | 0 | 2 | 2026-06-19 |
| 79 | Magistral | Mistral | 0 | 1 | 2026-08-03 |
| 80 | Magistral Medium 1 | Mistral | 0 | 2 | 2026-05-10 |
| 81 | Llama 4 Maverick | Meta | 0 | 1 | 2026-05-10 |
| 82 | Claude 3.7 Sonnet | Anthropic | 0 | 1 | 2026-05-10 |
| 83 | Qwen3 32B | Alibaba | 0 | 1 | 2026-06-15 |
| 84 | Grok 3 | SpaceXAI | 0 | 2 | 2026-06-19 |
| 85 | GPT-4.1 nano | OpenAI | 0 | 2 | 2026-06-19 |
| 86 | GPT-4o | OpenAI | 0 | 1 | 2026-05-10 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.