Calibration
GPQA Diamond → ARC-AGI-2
ARC-AGI-2 is estimated from GPQA Diamond with a linear curve fitted on 19 models measured on both: y = 5.4701·x + -4.3526, R² = 0.71, cross-validated error 9.8 pp. It is used for 46 estimates.
| Estimated model | GPQA Diamond | ARC-AGI-2 | Source |
|---|---|---|---|
| Sakana Fugu | 95.5% | 87.1% | estimated ± 9.8 pp, medium confidence |
| Sakana Fugu-Ultra | 95.5% | 87.1% | estimated ± 9.8 pp, medium confidence |
| Kimi K3.1 Pro | 94.2% | 80.0% | estimated ± 9.8 pp, medium confidence |
| Step 5 Preview | 93.5% | 76.2% | estimated ± 9.8 pp, medium confidence |
| Ornith-1.5-397B | 92.8% | 72.4% | estimated ± 9.8 pp, medium confidence |
| Qwen3.8 Max | 92.6% | 71.3% | estimated ± 9.8 pp, medium confidence |
| Pareto 26.10 Preview | 92.4% | 70.2% | estimated ± 9.8 pp, medium confidence |
| Qwen3.7 Max | 92.4% | 70.2% | estimated ± 9.8 pp, medium confidence |
| Hy4 preview | 92.3% | 69.6% | estimated ± 9.8 pp, medium confidence |
| Qwen3.8-Flash-Next | 91.7% | 66.3% | estimated ± 9.8 pp, medium confidence |
| GLM-5.2 | 91.2% | 63.6% | estimated ± 9.8 pp, medium confidence |
| Qwen3.8-Omni-Flash | 91.0% | 62.5% | estimated ± 9.8 pp, medium confidence |
| DeepSeek V4.1 Flash 1017 | 90.9% | 62.0% | estimated ± 9.8 pp, medium confidence |
| Beam 1.0 501B-A23B | 90.6% | 60.3% | estimated ± 9.8 pp, medium confidence |
| Kimi K2.6 | 90.5% | 59.8% | estimated ± 9.8 pp, medium confidence |
| Particle-1 | 90.4% | 59.2% | estimated ± 9.8 pp, medium confidence |
| Qwen3.7 Plus | 90.3% | 58.7% | estimated ± 9.8 pp, medium confidence |
| Beam 1.0 501B-A23B FP8 | 90.3% | 58.7% | estimated ± 9.8 pp, medium confidence |
| Interfaze Beta | 89.9% | 56.5% | estimated ± 9.8 pp, medium confidence |
| Nemotron 3.5 Super VL 120B A12B NVFP4 | 89.9% | 56.5% | estimated ± 9.8 pp, medium confidence |
| Nemotron 3.5 Super VL 120B A12B | 89.5% | 54.4% | estimated ± 9.8 pp, medium confidence |
| Beam 1.0 501B-A23B NVFP4 | 89.3% | 53.2% | estimated ± 9.8 pp, medium confidence |
| Mercury 3 | 89.2% | 52.8% | estimated ± 9.8 pp, medium confidence |
| Claude Opus 4.6 | 89.2% | 52.7% | estimated ± 9.8 pp, medium confidence |
| Ornith-1.5-35B-A3B | 89.2% | 52.7% | estimated ± 9.8 pp, medium confidence |
| Qwen3.8-27B | 89.2% | 52.7% | estimated ± 9.8 pp, medium confidence |
| Solar Pro 4 | 89.0% | 51.6% | estimated ± 9.8 pp, medium confidence |
| Nemotron 3.5 Bolt 74B-A9B BF16 | 88.8% | 50.6% | estimated ± 9.8 pp, medium confidence |
| OrcaCyber-Nano-1.0 | 88.3% | 47.6% | estimated ± 9.8 pp, medium confidence |
| Inkling | 87.9% | 45.6% | estimated ± 9.8 pp, low confidence |
| Kimi K2.5 | 87.6% | 43.9% | estimated ± 9.8 pp, low confidence |
| Hy3 Preview | 87.2% | 41.7% | estimated ± 9.8 pp, low confidence |
| MiniMax M2.7 | 87.0% | 40.6% | estimated ± 9.8 pp, low confidence |
| Nemotron 3 Ultra | 87.0% | 40.6% | estimated ± 9.8 pp, low confidence |
| Nemotron 3.5 Bolt 74B-A9B | 86.4% | 37.5% | estimated ± 9.8 pp, low confidence |
| Ornith-1.5-9B | 86.4% | 37.4% | estimated ± 9.8 pp, low confidence |
| Solar Open 2 | 86.3% | 36.8% | estimated ± 9.8 pp, low confidence |
| GLM-5.1 | 86.2% | 36.3% | estimated ± 9.8 pp, low confidence |
| GLM-5 | 86.0% | 35.2% | estimated ± 9.8 pp, low confidence |
| Ternary Bonsai 2 27B | 85.8% | 33.9% | estimated ± 9.8 pp, low confidence |
| A.X K2 | 85.6% | 33.0% | estimated ± 9.8 pp, low confidence |
| Ripple-65B Preview | 85.2% | 30.8% | estimated ± 9.8 pp, low confidence |
| Ling 3.0 Flash | 85.0% | 29.5% | estimated ± 9.8 pp, low confidence |
| MAI-Thinking-1 | 84.2% | 25.3% | estimated ± 9.8 pp, low confidence |
| Ling 3.0 Flash FP8 | 84.0% | 24.2% | estimated ± 9.8 pp, low confidence |
| K-EXAONE 2.0 | 82.2% | 14.4% | estimated ± 9.8 pp, low confidence |
Fitted on
| Model | GPQA Diamond | ARC-AGI-2 |
|---|---|---|
| Claude Opus 4.7 | 94.2% | 75.8% |
| Claude Opus 4.8 | 93.6% | 72.1% |
| DeepSeek V4 Flash 0731 | 88.1% | 61.4% |
| DeepSeek V4 Pro | 90.1% | 61.3% |
| Gemini 3.1 Pro | 94.3% | 77.1% |
| Gemini 3.5 Flash | 92.7% | 72.1% |
| GPT-5.4 | 92.8% | 74.0% |
| GPT-5.5 | 93.6% | 85.0% |
| GPT-5.6 Luna | 92.3% | 59.5% |
| GPT-5.6 Sol | 94.6% | 92.5% |
| GPT-5.6 Terra | 92.9% | 83.9% |
| GPT-6.2 Astra | 97.0% | 95.8% |
| GPT-6.2 Sol | 96.0% | 85.8% |
| GPT-6 Astra | 96.0% | 95.0% |
| Grok 4.20 | 88.5% | 53.3% |
| Inkling 2 Small | 93.0% | 81.7% |
| Inkling-Small | 89.5% | 40.1% |
| Kimi K3 | 93.5% | 60.4% |
| Muse Spark | 89.5% | 42.5% |