Calibration
Artificial Analysis Intelligence Index → GPQA
GPQA is estimated from Artificial Analysis Intelligence Index with a Hill curve fitted on 62 models measured on both: y = 0.0000 + (0.9382 − 0.0000)·x^2.30 / (0.07365^2.30 + x^2.30), R² = 0.89, cross-validated error 3.2 pp. It is used for 90 estimates.
| Estimated model | Artificial Analysis Intelligence Index | GPQA | Source |
|---|---|---|---|
| Apodex 1.1 Mini | 26.4% | 89.1% | estimated ± 3.2 pp, high confidence |
| Celeris-1 | 6.4% | 39.0% | estimated ± 3.2 pp, medium confidence |
| Claude 3 Haiku | 5.6% | 32.4% | estimated ± 3.2 pp, medium confidence |
| Claude 3 Opus | 8.7% | 55.9% | estimated ± 3.2 pp, high confidence |
| Claude 4.1 Opus | 18.6% | 83.8% | estimated ± 3.2 pp, high confidence |
| Claude 4.1 Opus Thinking | 22.9% | 87.3% | estimated ± 3.2 pp, high confidence |
| Claude 4 Sonnet | 16.6% | 81.3% | estimated ± 3.2 pp, high confidence |
| Command A+ | 22.5% | 87.1% | estimated ± 3.2 pp, high confidence |
| DeepSeek-R1 | 13.1% | 74.2% | estimated ± 3.2 pp, high confidence |
| DeepSeek R1 Distill Qwen 32B | 8.4% | 53.8% | estimated ± 3.2 pp, high confidence |
| DeepSeek V3 0324 | 9.7% | 61.4% | estimated ± 3.2 pp, high confidence |
| DeepSeek V3.1 | 13.7% | 75.7% | estimated ± 3.2 pp, high confidence |
| DeepSeek V3.1 (Reasoning) | 13.5% | 75.1% | estimated ± 3.2 pp, high confidence |
| DeepSeek V3.2 | 16.0% | 80.4% | estimated ± 3.2 pp, high confidence |
| Exaone 4.0 1.2B | 5.2% | 29.3% | estimated ± 3.2 pp, medium confidence |
| Exaone 4.0 32B | 6.3% | 38.8% | estimated ± 3.2 pp, medium confidence |
| Gemini 1.0 Pro | 5.3% | 30.3% | estimated ± 3.2 pp, medium confidence |
| Gemini 1.5 Pro | 7.9% | 50.7% | estimated ± 3.2 pp, high confidence |
| Gemini 2.5 Flash | 9.9% | 62.0% | estimated ± 3.2 pp, high confidence |
| Gemini 4 Argon | 52.6% | 92.8% | estimated ± 3.2 pp, high confidence |
| Gemma 3 27B | 4.9% | 26.0% | estimated ± 3.2 pp, medium confidence |
| GLM-4.5-Air | 11.1% | 67.5% | estimated ± 3.2 pp, high confidence |
| GLM-5-Turbo | 26.6% | 89.2% | estimated ± 3.2 pp, high confidence |
| GLM-5V-Turbo | 23.5% | 87.7% | estimated ± 3.2 pp, high confidence |
| GPT-4 Turbo | 7.0% | 44.5% | estimated ± 3.2 pp, medium confidence |
| GPT-4o | 8.4% | 54.2% | estimated ± 3.2 pp, high confidence |
| GPT-4o mini | 6.7% | 41.5% | estimated ± 3.2 pp, medium confidence |
| GPT-5.1-Codex | 23.7% | 87.8% | estimated ± 3.2 pp, high confidence |
| GPT-5.1-Codex-Max | 23.7% | 87.8% | estimated ± 3.2 pp, high confidence |
| GPT-5.2-Codex | 28.5% | 89.8% | estimated ± 3.2 pp, high confidence |
| GPT-5.3 Codex | 32.5% | 90.8% | estimated ± 3.2 pp, high confidence |
| GPT-5 (high) | 23.0% | 87.4% | estimated ± 3.2 pp, high confidence |
| GPT-5 (medium) | 22.9% | 87.4% | estimated ± 3.2 pp, high confidence |
| GPT-OSS 120B | 11.6% | 69.4% | estimated ± 3.2 pp, high confidence |
| GPT-OSS 20B | 9.0% | 57.4% | estimated ± 3.2 pp, high confidence |
| Granite-4.0-350M | 4.8% | 25.8% | estimated ± 3.2 pp, medium confidence |
| Granite-4.0-H-1B | 5.2% | 29.0% | estimated ± 3.2 pp, medium confidence |
| Granite-4.0-H-350M | 4.8% | 25.8% | estimated ± 3.2 pp, medium confidence |
| Grok 4 | 22.5% | 87.1% | estimated ± 3.2 pp, high confidence |
| Grok 4.1 Fast | 11.3% | 68.2% | estimated ± 3.2 pp, high confidence |
| Grok 4.1 Fast (Reasoning) | 20.4% | 85.6% | estimated ± 3.2 pp, high confidence |
| Grok 4 Fast (Reasoning) | 17.9% | 83.1% | estimated ± 3.2 pp, high confidence |
| Grok Code Fast 1 | 14.1% | 76.5% | estimated ± 3.2 pp, high confidence |
| Hy3 | 25.3% | 88.6% | estimated ± 3.2 pp, high confidence |
| K-Exaone | 14.4% | 77.2% | estimated ± 3.2 pp, high confidence |
| Kimi K2 | 12.7% | 73.0% | estimated ± 3.2 pp, high confidence |
| Kimi K2.7 Code | 25.8% | 88.8% | estimated ± 3.2 pp, high confidence |
| LFM2.5-2.6B | 8.4% | 53.9% | estimated ± 3.2 pp, high confidence |
| LFM2.5-8B-A1B | 7.2% | 45.8% | estimated ± 3.2 pp, medium confidence |
| LFM2.5-VL-1.6B-Extract | 4.8% | 25.8% | estimated ± 3.2 pp, medium confidence |
| Ling 3.0 Flash VL | 24.6% | 88.3% | estimated ± 3.2 pp, high confidence |
| Ling 3.0 Tiny | 11.1% | 67.4% | estimated ± 3.2 pp, high confidence |
| Llama 3.1 405B | 7.3% | 46.4% | estimated ± 3.2 pp, medium confidence |
| Llama 4 Maverick | 10.0% | 62.7% | estimated ± 3.2 pp, high confidence |
| Llama 4 Scout | 8.1% | 51.9% | estimated ± 3.2 pp, high confidence |
| MiMo-V2.6-Flash | 37.9% | 91.7% | estimated ± 3.2 pp, high confidence |
| MiMo-V2.6-Pro | 46.3% | 92.5% | estimated ± 3.2 pp, high confidence |
| MiMo-V2-Omni | 23.9% | 88.0% | estimated ± 3.2 pp, high confidence |
| MiMo-V2-Pro | 28.6% | 89.9% | estimated ± 3.2 pp, high confidence |
| Mistral Large 2 | 7.6% | 48.3% | estimated ± 3.2 pp, medium confidence |
| Mistral Large 3 | 9.3% | 59.0% | estimated ± 3.2 pp, high confidence |
| Mistral Large 4 | 38.4% | 91.8% | estimated ± 3.2 pp, high confidence |
| Mistral Medium 3 | 9.0% | 57.8% | estimated ± 3.2 pp, high confidence |
| Mistral Small 4 | 11.3% | 68.2% | estimated ± 3.2 pp, high confidence |
| Mistral Small 4 (Reasoning) | 11.3% | 68.2% | estimated ± 3.2 pp, high confidence |
| Muse Glimmer 30B | 17.5% | 82.5% | estimated ± 3.2 pp, high confidence |
| Muse Spark 1.3 | 48.1% | 92.6% | estimated ± 3.2 pp, high confidence |
| Nemotron 3 Nano 30B | 8.9% | 57.0% | estimated ± 3.2 pp, high confidence |
| Nemotron 3 Super 100B | 12.8% | 73.3% | estimated ± 3.2 pp, high confidence |
| Nemotron Ultra 253B | 7.5% | 48.1% | estimated ± 3.2 pp, medium confidence |
| North Mini Code | 9.9% | 62.3% | estimated ± 3.2 pp, high confidence |
| Nova Pro | 7.0% | 43.9% | estimated ± 3.2 pp, medium confidence |
| o1-preview | 11.4% | 68.6% | estimated ± 3.2 pp, high confidence |
| o3 | 20.2% | 85.4% | estimated ± 3.2 pp, high confidence |
| o3-pro | 21.9% | 86.7% | estimated ± 3.2 pp, high confidence |
| Phi-4 | 5.9% | 35.4% | estimated ± 3.2 pp, medium confidence |
| Phi-4 Multimodal Instruct | 5.8% | 34.4% | estimated ± 3.2 pp, medium confidence |
| Quasar 438B | 26.7% | 89.2% | estimated ± 3.2 pp, high confidence |
| Qwen2.5 Coder 32B Instruct | 6.7% | 42.1% | estimated ± 3.2 pp, medium confidence |
| Qwen3.5 397B (Reasoning) | 21.4% | 86.4% | estimated ± 3.2 pp, high confidence |
| Qwen3.8 Max Preview | 45.4% | 92.4% | estimated ± 3.2 pp, high confidence |
| Qwen3 Max | 15.6% | 79.7% | estimated ± 3.2 pp, high confidence |
| Qwen3-Omni-30B-A3B-Instruct | 6.0% | 36.1% | estimated ± 3.2 pp, medium confidence |
| Qwen3-Omni-30B-A3B-Thinking | 7.8% | 49.7% | estimated ± 3.2 pp, medium confidence |
| Sarvam 105B | 8.8% | 56.3% | estimated ± 3.2 pp, high confidence |
| Sarvam 30B | 6.6% | 40.7% | estimated ± 3.2 pp, medium confidence |
| Solar Pro 2 | 7.0% | 44.2% | estimated ± 3.2 pp, medium confidence |
| Solar Pro 3 | 7.8% | 50.1% | estimated ± 3.2 pp, high confidence |
| Step 3.7 Flash | 19.5% | 84.8% | estimated ± 3.2 pp, high confidence |
| Ultravox v0.6 Llama 3.3 70B | 7.7% | 49.0% | estimated ± 3.2 pp, medium confidence |