Benchmarks
ARC-AGI
Models scored
159
Best score
0.98
Metric
Score
ARC-AGI is an AI evaluation tracked by Epoch AI, with 159 scored model versions on record. The scores here are reported by third parties rather than produced by Epoch AI running the evaluation itself.
Scored models span September 2024 to July 2026. The best score recorded is 0.98, measured as Score. Models come from Google DeepMind, OpenAI, Anthropic, Google, xAI, Z.ai (Zhipu AI), Moonshot, MiniMax and others.
Full record
- Score metric
- Score
- Models scored
- 159
- Best score recorded
- 0.98
- Earliest model scored
- Sep 12, 2024
- Most recent model scored
- Jul 24, 2026
- Third-party reported
- Yes
Leaderboard
| 01 | gemini-3.1-pro-preview | Google DeepMind | 0.98 |
| 02 | claude-opus-5_max | Anthropic | 0.97 |
| 03 | gpt-5.6-sol_xhigh | OpenAI | 0.97 |
| 04 | gpt-5.6-sol_high | OpenAI | 0.97 |
| 05 | gpt-5.5-pro_high | OpenAI | 0.96 |
| 06 | gpt-5.6-sol_max | OpenAI | 0.96 |
| 07 | gpt-5.6-terra_max | OpenAI | 0.96 |
| 08 | gpt-5.5_xhigh | OpenAI | 0.95 |
| 09 | gpt-5.5-pro_xhigh | OpenAI | 0.95 |
| 10 | gpt-5.4-pro-2026-03-05_xhigh | OpenAI | 0.95 |
| 11 | gpt-5.5_high | OpenAI | 0.95 |
| 12 | gpt-5.6-terra_xhigh | OpenAI | 0.94 |
| 13 | claude-opus-4-6_120K | Anthropic | 0.94 |
| 14 | gpt-5.4-2026-03-05_xhigh | OpenAI | 0.94 |
| 15 | claude-opus-4-7_high | Anthropic | 0.94 |
| 16 | gpt-5.4-2026-03-05_high | OpenAI | 0.93 |
| 17 | gemini-3.5-flash_high | 0.93 | |
| 18 | claude-opus-4-8_max | Anthropic | 0.93 |
| 19 | gpt-5.6-sol_medium | OpenAI | 0.93 |
| 20 | gpt-5.5_medium | OpenAI | 0.92 |
| 21 | claude-opus-4-8_high | Anthropic | 0.92 |
| 22 | claude-opus-4-7_max | Anthropic | 0.92 |
| 23 | gpt-5.6-terra_high | OpenAI | 0.92 |
| 24 | claude-opus-4-8_medium | Anthropic | 0.92 |
| 25 | claude-opus-4-7_medium | Anthropic | 0.91 |
| 26 | claude-opus-4-7_low | Anthropic | 0.91 |
| 27 | gpt-5.2-pro-2025-12-11_xhigh | OpenAI | 0.91 |
| 28 | grok-4-20 | xAI | 0.9 |
| 29 | claude-opus-4-8_low | Anthropic | 0.88 |
| 30 | gpt-5.6-luna_max | OpenAI | 0.88 |
| 31 | gpt-5.6-luna_xhigh | OpenAI | 0.88 |
| 32 | gemini-3-deep-think-preview | 0.88 | |
| 33 | claude-sonnet-4-6_high | Anthropic | 0.86 |
| 34 | gpt-5.2-2025-12-11_xhigh | OpenAI | 0.86 |
| 35 | gpt-5.4-2026-03-05_medium | OpenAI | 0.86 |
| 36 | claude-sonnet-4-6_max | Anthropic | 0.86 |
| 37 | gpt-5.2-pro-2025-12-11_high | OpenAI | 0.86 |
| 38 | gpt-5.2-pro-2025-12-11_medium | OpenAI | 0.81 |
| 39 | claude-opus-4-5-20251101_64K | Anthropic | 0.8 |
| 40 | gpt-5.2-2025-12-11_high | OpenAI | 0.79 |
| 41 | gpt-5.6-terra_medium | OpenAI | 0.77 |
| 42 | glm-5.2_unknown | Z.ai (Zhipu AI) | 0.77 |
| 43 | gpt-5.6-luna_high | OpenAI | 0.77 |
| 44 | gpt-5.5_low | OpenAI | 0.76 |
| 45 | claude-opus-4-5-20251101_32K | Anthropic | 0.76 |
| 46 | gemini-3-pro-preview | Google DeepMind | 0.75 |
| 47 | gpt-5.6-sol_low | OpenAI | 0.74 |
| 48 | gpt-5.1-2025-11-13_high | OpenAI | 0.73 |
| 49 | gpt-5.2-2025-12-11_medium | OpenAI | 0.73 |
| 50 | claude-opus-4-5-20251101_16K | Anthropic | 0.72 |
| 51 | gpt-5-pro-2025-10-06_high | OpenAI | 0.7 |
| 52 | gpt-5-pro-2025-10-06_unknown | OpenAI | 0.7 |
| 53 | gpt-5.4-2026-03-05_low | OpenAI | 0.68 |
| 54 | grok-4-0709 | xAI | 0.67 |
| 55 | gpt-5-2025-08-07_high | OpenAI | 0.66 |
| 56 | kimi-k2.5 | Moonshot | 0.65 |
| 57 | MiniMax-M2.5 | MiniMax | 0.64 |
| 58 | claude-sonnet-4-5-20250929_32K | Anthropic | 0.64 |
| 59 | gpt-5.4-mini-2026-03-17_xhigh | OpenAI | 0.64 |
| 60 | o3-2025-04-16_high | OpenAI | 0.61 |
More benchmarks
SourceEpoch AI, 'AI Benchmarking Hub'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/benchmarks. Licensed under CC BY 4.0.
← All benchmarks