Benchmark
ARC-AGI 2
Scores are percentages. Higher is better.
- Models measured
- 98
- Scores recorded
- 318
- Sources
- 3
- IQ weight
- 1.29
Score distribution
- 42.4%
- Mean
- 42.5%
- Median
- 35%
- Std dev
- 0.9%
- 10th percentile
- 90%
- 90th percentile
- 0%
- Min
This benchmark has a weight of 1.29 for chat models, measured over 98 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 51 | GPT-6.2 Solhigh· Epoch AI Benchmarking Hub data | 85.8% |
| 52 | GPT-6.2 Solmax· Epoch AI Benchmarking Hub data | 85.8% |
| 53 | GPT-6.2 Solxhigh· Epoch AI Benchmarking Hub data | 85.8% |
| 54 | GPT-5.6 Solhigh· Epoch AI Benchmarking Hub data | 85.4% |
| 55 | GPT-6 Astralow· Epoch AI Benchmarking Hub data | 85.4% |
| 56 | GPT 5.5Verified · xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 85% |
| 57 | GPT 5.5xhigh· Epoch AI Benchmarking Hub data | 85% |
| 58 | GPT 5.5Gemini 3.5 Flash Model Card · May 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 84.6% |
| 59 | Gemini 3.7 Flashhigh· Epoch AI Benchmarking Hub data | 84.6% |
| 60 | GPT-5.5 Prohigh· Epoch AI Benchmarking Hub data | 84.6% |
| 61 | GPT-5.5 Proxhigh· Epoch AI Benchmarking Hub data | 84.2% |
| 62 | GPT-5.6 Terramax· Epoch AI Benchmarking Hub data | 83.9% |
| 63 | Claude Sonnet 5.5xhigh· Epoch AI Benchmarking Hub data | 83.9% |
| 64 | GPT 5.5high· Epoch AI Benchmarking Hub data | 83.3% |
| 65 | GPT-5.4 Pro· Epoch AI Benchmarking Hub data | 83.3% |
| 66 | GPT-5.4 ProVerified· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 83.3% |
| 67 | Gemini 3.8 Flashmedium· Epoch AI Benchmarking Hub data | 82.9% |
| 68 | GPT-6.2 Solmedium· Epoch AI Benchmarking Hub data | 82.9% |
| 69 | Claude Fable 5medium· Epoch AI Benchmarking Hub data | 82.5% |
| 70 | Inkling 2 Smallxhigh· Epoch AI Benchmarking Hub data | 81.7% |
| 71 | Claude Fable 5.1low· Epoch AI Benchmarking Hub data | 78.3% |
| 72 | DeepSeek V4.1 Flash 1017max· Epoch AI Benchmarking Hub data | 78.1% |
| 73 | GPT-6 Solxhigh· Epoch AI Benchmarking Hub data | 78.1% |
| 74 | Gemini 3.8 Flashlow· Epoch AI Benchmarking Hub data | 77.5% |
| 75 | Gemini 3.1 Pro Preview· Epoch AI Benchmarking Hub data | 77.1% |
| 76 | Gemini 3.1 Pro PreviewARC Prize Verified· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 77.1% |
| 77 | Gemini 3.1 Pro PreviewGemini 3.5 Flash Model Card · May 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 77.1% |
| 78 | Claude Fable 5low· Epoch AI Benchmarking Hub data | 76.8% |
| 79 | GPT-6.1 Sollow· Epoch AI Benchmarking Hub data | 76.7% |
| 80 | Claude Opus 4.7max· Epoch AI Benchmarking Hub data | 75.8% |
| 81 | Claude Opus 4.7private validation · adaptive thinking · max effort · capability summary· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 75.8% |
| 82 | Claude Opus 4.7Gemini 3.5 Flash Model Card · May 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 75.8% |
| 83 | GPT-5.6 Terraxhigh· Epoch AI Benchmarking Hub data | 74.2% |
| 84 | GPT 5.4xhigh· Epoch AI Benchmarking Hub data | 74% |
| 85 | DeepSeek V4.1 Flash 1017low· Epoch AI Benchmarking Hub data | 73.6% |
| 86 | GPT 5.4Verified· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 73.3% |
| 87 | DeepSeek V4.1 Flashmax· Epoch AI Benchmarking Hub data | 72.9% |
| 88 | Claude Opus 4.8Opus 5 System Card capability summary · adaptive thinking · max effort unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 72.1% |
| 89 | Gemini 3.5 FlashGemini 3.5 Flash Model Card · May 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 72.1% |
| 90 | Claude Opus 4.8high· Epoch AI Benchmarking Hub data | 72.1% |
| 91 | Gemini 3.5 Flashhigh· Epoch AI Benchmarking Hub data | 72.1% |
| 92 | Claude Opus 4.8medium· Epoch AI Benchmarking Hub data | 71.7% |
| 93 | GPT 5.5medium· Epoch AI Benchmarking Hub data | 70.4% |
| 94 | Claude Opus 5.5low· Epoch AI Benchmarking Hub data | 70.1% |
| 95 | Claude Opus 4.6120k· Epoch AI Benchmarking Hub data | 69.2% |
| 96 | Claude Opus 4.6120K thinking tokens · high effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 69.2% |
| 97 | GPT-6 Solhigh· Epoch AI Benchmarking Hub data | 68.9% |
| 98 | Claude Opus 4.6Opus 4.6 System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials unless noted · Verified· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 68.8% |
| 99 | DeepSeek V4.1 Flash 1017high· Epoch AI Benchmarking Hub data | 68.6% |
| 100 | Claude Opus 4.7high· Epoch AI Benchmarking Hub data | 68.3% |
Showing 51–100 of 318 results.
Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · Data by LLM Stats