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 |
|---|---|---|
| 1 | GPT-6.2 Astramax· Epoch AI Benchmarking Hub data | 95.8% |
| 2 | GPT-6.2 Astraxhigh· Epoch AI Benchmarking Hub data | 95.8% |
| 3 | GPT-6 Astramax· Epoch AI Benchmarking Hub data | 95% |
| 4 | GPT-6 AstraOpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 95% |
| 5 | GPT-6.2 Astrahigh· Epoch AI Benchmarking Hub data | 94.6% |
| 6 | Gemini 4 Carbonmedium· Epoch AI Benchmarking Hub data | 94.2% |
| 7 | GPT-6.1 Solmax· Epoch AI Benchmarking Hub data | 94.2% |
| 8 | Claude Fable 5.2xhigh· Epoch AI Benchmarking Hub data | 93.8% |
| 9 | Claude Opus 5.5high· Epoch AI Benchmarking Hub data | 93.3% |
| 10 | GPT-6 Astraxhigh· Epoch AI Benchmarking Hub data | 93.3% |
| 11 | GPT-6.2 Astramedium· Epoch AI Benchmarking Hub data | 93.3% |
| 12 | Claude Fable 5.2max· Epoch AI Benchmarking Hub data | 92.5% |
| 13 | Claude Opus 5.5xhigh· Epoch AI Benchmarking Hub data | 92.5% |
| 14 | GPT-5.6 Solmax· Epoch AI Benchmarking Hub data | 92.5% |
| 15 | GPT-5.6 SolOpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 92.5% |
| 16 | GPT-5.6 SolOpus 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. | 92.5% |
| 17 | GPT-5.6 SolSystem Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 92.5% |
| 18 | Gemini 4 Carbonhigh· Epoch AI Benchmarking Hub data | 92.1% |
| 19 | GPT-6 Astrahigh· Epoch AI Benchmarking Hub data | 92.1% |
| 20 | GPT-6 Astramedium· Epoch AI Benchmarking Hub data | 92.1% |
| 21 | Claude Opus 5.5max· Epoch AI Benchmarking Hub data | 91.7% |
| 22 | GPT-6.1 Solhigh· Epoch AI Benchmarking Hub data | 91.7% |
| 23 | GPT-6.1 Solxhigh· Epoch AI Benchmarking Hub data | 91.7% |
| 24 | GPT-6.2 Astralow· Epoch AI Benchmarking Hub data | 91.7% |
| 25 | Gemini 4 Argonhigh· Epoch AI Benchmarking Hub data | 91.3% |
| 26 | Claude Opus 5max· Epoch AI Benchmarking Hub data | 90.4% |
| 27 | Claude Opus 5System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 90.4% |
| 28 | Claude Fable 5.2high· Epoch AI Benchmarking Hub data | 90.4% |
| 29 | Claude Opus 5OpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 90.4% |
| 30 | Claude Opus 5Opus 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. | 90.4% |
| 31 | Claude Fable 5.1max· Epoch AI Benchmarking Hub data | 90% |
| 32 | Claude Fable 5.1OpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 90% |
| 33 | Claude Fable 5.1System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 90% |
| 34 | Claude Fable 5.1xhigh· Epoch AI Benchmarking Hub data | 90% |
| 35 | GPT-5.6 Solxhigh· Epoch AI Benchmarking Hub data | 90% |
| 36 | Claude Fable 5.2medium· Epoch AI Benchmarking Hub data | 89.6% |
| 37 | GPT-6 Solmax· Epoch AI Benchmarking Hub data | 89.6% |
| 38 | Claude Fable 5OpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 89.2% |
| 39 | Claude Fable 5System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 89.2% |
| 40 | Claude Fable 5max· Epoch AI Benchmarking Hub data | 89.2% |
| 41 | Gemini 3.8 Flashhigh· Epoch AI Benchmarking Hub data | 89.2% |
| 42 | Claude Fable 5.1high· Epoch AI Benchmarking Hub data | 88.8% |
| 43 | Claude Fable 5xhigh· Epoch AI Benchmarking Hub data | 88.3% |
| 44 | Claude Opus 5high· Epoch AI Benchmarking Hub data | 88.3% |
| 45 | Claude Fable 5high· Epoch AI Benchmarking Hub data | 87.5% |
| 46 | Claude Opus 5.5medium· Epoch AI Benchmarking Hub data | 87.5% |
| 47 | Gemini 4 Argonmedium· Epoch AI Benchmarking Hub data | 87.1% |
| 48 | Claude Fable 5.2low· Epoch AI Benchmarking Hub data | 86.7% |
| 49 | GPT-6.1 Solmedium· Epoch AI Benchmarking Hub data | 86.7% |
| 50 | Claude Fable 5.1medium· Epoch AI Benchmarking Hub data | 86.3% |
Showing 1–50 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