Calibration
Artificial Analysis Intelligence Index → MMLU-Pro (Arcee)
MMLU-Pro (Arcee) is estimated from Artificial Analysis Intelligence Index with a linear curve fitted on 6 models measured on both: y = 0.4676·x + 0.7405, R² = 0.50, cross-validated error 5.4 pp. It is used for 13 estimates.
| Estimated model | Artificial Analysis Intelligence Index | MMLU-Pro (Arcee) | Source |
|---|---|---|---|
| Claude 3 Opus | 8.7% | 78.1% | estimated ± 5.4 pp, low confidence |
| Claude 4.1 Opus | 18.6% | 82.7% | estimated ± 5.4 pp, low confidence |
| DeepSeek R1 Distill Qwen 32B | 8.4% | 78.0% | estimated ± 5.4 pp, low confidence |
| Gemini 1.0 Pro | 5.3% | 76.5% | estimated ± 5.4 pp, low confidence |
| Gemini 1.5 Pro | 7.9% | 77.7% | estimated ± 5.4 pp, low confidence |
| GPT-4 Turbo | 7.0% | 77.3% | estimated ± 5.4 pp, low confidence |
| GPT-4o mini | 6.7% | 77.2% | estimated ± 5.4 pp, low confidence |
| o1-preview | 11.4% | 79.4% | estimated ± 5.4 pp, low confidence |
| o1-pro | 12.4% | 79.9% | estimated ± 5.4 pp, low confidence |
| o3-mini | 12.5% | 79.9% | estimated ± 5.4 pp, low confidence |
| o3-pro | 21.9% | 84.3% | estimated ± 5.4 pp, low confidence |
| Phi-4 Multimodal Instruct | 5.8% | 76.8% | estimated ± 5.4 pp, low confidence |
| Qwen2.5 Coder 32B Instruct | 6.7% | 77.2% | estimated ± 5.4 pp, low confidence |