AI2 ARC-Challenge — Classic AI2 Reasoning Challenge (science QA) — unrelated to ARC-AGI despite shared 'ARC' acronym
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | Nemotron 3 Ultra 550B A55B Base | NVIDIA | 97.3 | 1 | 2026-06-05 |
| 2 | NVIDIA-Nemotron-3-Ultra-550B-A55B-Base-BF16 | NVIDIA | 97.3 | 1 | 2026-08-24 |
| 3 | Mistral Large 3 675B Base 2512 | Mistral | 97.3 | 1 | 2026-06-05 |
| 4 | MiMo V2.5 Pro Base | Xiaomi | 97.2 | 3 | 2026-06-05 |
| 5 | MiMo V2.5 Pro | Xiaomi | 97.2 | 1 | 2026-08-23 |
| 6 | Llama 3.1 Instruct 405B | Meta | 96.9 | 4 | 2026-08-23 |
| 7 | Kimi K2 Base | Moonshot | 96.7 | 8 | 2026-06-13 |
| 8 | GLM-4.5-Base | Z.ai | 96.6 | 3 | 2026-06-05 |
| 9 | MiMo V2.5 Base | Xiaomi | 96.5 | 3 | 2026-06-05 |
| 10 | Claude 3 Opus | Anthropic | 96.4 | 1 | 2026-08-23 |
| 11 | Nemotron 3 Super 120B A12B | NVIDIA | 96.1 | 1 | 2026-08-11 |
| 12 | Hy3 preview-Base | Tencent | 96 | 2 | 2026-06-05 |
| 13 | MiMo V2 Flash Base | Xiaomi | 95.9 | 2 | 2026-06-13 |
| 14 | DeepSeek-V3.1-Base | DeepSeek | 95.6 | 2 | 2026-06-13 |
| 15 | DeepSeek-V3.2-Exp-Base | DeepSeek | 95.5 | 3 | 2026-06-13 |
| 16 | Qwen3.5 35B A3B | Alibaba | 95.4 | 1 | 2026-08-11 |
| 17 | DeepSeek-V3 | DeepSeek | 95.3 | 1 | 2026-05-01 |
| 18 | Hunyuan Large | Tencent | 95 | 1 | 2026-05-31 |
| 19 | Llama 3.1 70B Instruct | Meta | 94.8 | 2 | 2026-08-23 |
| 20 | Nova Pro (Non-Reasoning) | Amazon | 94.8 | 1 | 2026-08-23 |
| 21 | DeepSeek-V3-Base | DeepSeek | 94.7 | 2 | 2026-06-05 |
| 22 | Hunyuan Large Inst. | Tencent | 94.6 | 1 | 2026-05-31 |
| 23 | Qwen2.5 Instruct 72B | Alibaba | 94.5 | 1 | 2026-05-01 |
| 24 | Claude 3 Sonnet | Anthropic | 93.2 | 1 | 2026-08-23 |
| 25 | Jamba 1.5 Large | AI21 | 93 | 1 | 2026-08-23 |
| 26 | Gemma 4 26B A4B | 92.8 | 1 | 2026-08-11 | |
| 27 | NVIDIA-Nemotron-3.5-Lightning-30B-A3B-Base-BF16 | NVIDIA | 92.7 | 1 | 2026-08-24 |
| 28 | Nemotron 3.5 Lightning | NVIDIA | 92.7 | 1 | 2026-08-11 |
| 29 | Nova Lite (Non-Reasoning) | Amazon | 92.4 | 1 | 2026-08-23 |
| 30 | DeepSeek-V2 | DeepSeek | 92.4 | 2 | 2026-05-31 |
| 31 | NVIDIA Nemotron 3 Nano 30B A3B | NVIDIA | 92 | 1 | 2026-08-11 |
| 32 | Phi 3.5 MoE Instruct | Microsoft | 91 | 1 | 2026-08-23 |
| 33 | Claude 2 | Anthropic | 91 | 1 | 2026-07-17 |
| 34 | Nova Micro (Non-Reasoning) | Amazon | 90.2 | 1 | 2026-08-23 |
| 35 | Claude 1.3 | Anthropic | 90 | 1 | 2026-07-17 |
| 36 | Claude 3 Haiku | Anthropic | 89.2 | 1 | 2026-08-23 |
| 37 | Jamba 1.5 Mini | AI21 | 85.7 | 1 | 2026-08-23 |
| 38 | Phi 3.5 Mini Instruct | Microsoft | 84.6 | 1 | 2026-08-23 |
| 39 | Phi 4 Mini Instruct | Microsoft | 83.7 | 1 | 2026-08-23 |
| 40 | Llama 3.1 8B Instruct | Meta | 83.4 | 1 | 2026-08-23 |
| 41 | Llama 3.2 3B Instruct | Meta | 78.6 | 1 | 2026-08-23 |
| 42 | Gemma 2 27B IT | 71.4 | 1 | 2026-08-23 | |
| 43 | Command R+ (Apr '24) | Cohere | 71 | 1 | 2026-08-23 |
| 44 | Qwen 2.5 Coder 32B Instruct | Alibaba | 70.5 | 1 | 2026-08-23 |
| 45 | Llama 3.1 Nemotron Instruct 70B | NVIDIA | 69.2 | 1 | 2026-08-23 |
| 46 | Qwen2 Instruct (72B) | Alibaba | 68.9 | 1 | 2026-08-23 |
| 47 | Gemma 2 Instruct (9B) | 68.4 | 1 | 2026-08-23 | |
| 48 | ERNIE 4.5 | Baidu | 40.6 | 1 | 2026-08-23 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.