Calibration
GDPval-AA → HLE w/ tools
HLE w/ tools is estimated from GDPval-AA with a linear curve fitted on 13 models measured on both: y = 0.5134·x + 0.3158, R² = 0.64, cross-validated error 6.1 pp. It is used for 72 estimates.
| Estimated model | GDPval-AA | HLE w/ tools | Source |
|---|---|---|---|
| A.X K2 | 23.2% | 43.5% | estimated ± 6.1 pp, low confidence |
| Apodex 1.1 Mini | 34.8% | 49.4% | estimated ± 6.1 pp, medium confidence |
| Celeris-1 | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Command A+ | 0.9% | 32.0% | estimated ± 6.1 pp, low confidence |
| DeepSeek V3 | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| DeepSeek V3 0324 | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Gemini 2.5 Pro | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Gemini 3.1 Pro | 14.7% | 39.1% | estimated ± 6.1 pp, low confidence |
| Gemini 3.5 Flash | 42.2% | 53.2% | estimated ± 6.1 pp, medium confidence |
| Gemini 3.5 Flash-Lite | 24.3% | 44.1% | estimated ± 6.1 pp, low confidence |
| Gemini 3.6 Flash | 39.3% | 51.8% | estimated ± 6.1 pp, medium confidence |
| Gemini 4 Argon | 56.3% | 60.5% | estimated ± 6.1 pp, medium confidence |
| Gemma 3 27B | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Gemma 4 12B | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Gemma 4 26B A4B | 3.4% | 33.3% | estimated ± 6.1 pp, low confidence |
| Gemma 4 31B | 6.1% | 34.7% | estimated ± 6.1 pp, low confidence |
| Gemma 4 E2B | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Gemma 4 E4B | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| GPT-4.1 mini | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| GPT-4.1 nano | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| GPT-4o | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| GPT-4o mini | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| GPT-5.1 | 16.5% | 40.1% | estimated ± 6.1 pp, low confidence |
| GPT-5.4 mini | 25.8% | 44.8% | estimated ± 6.1 pp, medium confidence |
| GPT-5.4 nano | 22.5% | 43.1% | estimated ± 6.1 pp, low confidence |
| GPT-5 (high) | 21.0% | 42.4% | estimated ± 6.1 pp, low confidence |
| GPT-6.1 Sol | 53.8% | 59.2% | estimated ± 6.1 pp, medium confidence |
| GPT-6 Luna | 46.9% | 55.7% | estimated ± 6.1 pp, medium confidence |
| GPT-OSS 120B | 5.6% | 34.5% | estimated ± 6.1 pp, low confidence |
| GPT-OSS 20B | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Granite 4.2 30B | 3.9% | 33.6% | estimated ± 6.1 pp, low confidence |
| Granite 4.2 3B | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Granite 4.2 8B | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Grok 4.3 | 29.2% | 46.6% | estimated ± 6.1 pp, medium confidence |
| Grok 4.5 | 44.5% | 54.4% | estimated ± 6.1 pp, medium confidence |
| Grok 4.6 | 56.1% | 60.4% | estimated ± 6.1 pp, medium confidence |
| Hy3 | 28.3% | 46.1% | estimated ± 6.1 pp, medium confidence |
| Hy3 Preview | 35.8% | 50.0% | estimated ± 6.1 pp, medium confidence |
| K-Exaone | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Kimi K2.7 Code | 27.0% | 45.4% | estimated ± 6.1 pp, medium confidence |
| LFM2.5-2.6B | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Ling 2.6 Flash | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Ling 3.0 Flash FP8 | 22.4% | 43.1% | estimated ± 6.1 pp, low confidence |
| Ling 3.0 Flash VL | 33.2% | 48.6% | estimated ± 6.1 pp, medium confidence |
| Ling 3.0 Tiny | 3.2% | 33.2% | estimated ± 6.1 pp, low confidence |
| Llama 4 Maverick | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Llama 4 Scout | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Mercury 2.5 | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| MiMo-V2.5-Pro | 31.2% | 47.6% | estimated ± 6.1 pp, medium confidence |
| MiMo-V2-Flash | 6.6% | 35.0% | estimated ± 6.1 pp, low confidence |
| MiniCPM5-2B | 10.5% | 37.0% | estimated ± 6.1 pp, low confidence |
| MiniMax M2.7 | 25.7% | 44.8% | estimated ± 6.1 pp, low confidence |
| Mistral Large 3 | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Mistral Large 4 | 46.2% | 55.3% | estimated ± 6.1 pp, medium confidence |
| Mistral Medium 3.5 128B | 13.2% | 38.4% | estimated ± 6.1 pp, low confidence |
| Mistral Small 4 | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Mistral Small 4 (Reasoning) | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Muse Glimmer 30B | 14.5% | 39.0% | estimated ± 6.1 pp, low confidence |
| Muse Spark 1.2 | 48.9% | 56.7% | estimated ± 6.1 pp, medium confidence |
| Nemotron 3 Nano 30B | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Nemotron 3 Nano Omni 30B A3B | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Nemotron 3 Super 100B | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| North Mini Code | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Quasar 438B | 34.2% | 49.1% | estimated ± 6.1 pp, medium confidence |
| Qwen3.6-27B | 24.4% | 44.1% | estimated ± 6.1 pp, low confidence |
| Qwen3.7 Plus | 13.5% | 38.5% | estimated ± 6.1 pp, low confidence |
| Qwen3.8-27B | 46.2% | 55.3% | estimated ± 6.1 pp, medium confidence |
| Qwen3.8 Max Preview | 58.6% | 61.7% | estimated ± 6.1 pp, medium confidence |
| Solar Pro 3 | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Trinity-Large-Preview | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Trinity-Large-Thinking | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |
| Ultravox v0.6 Llama 3.3 70B | 0.0% | 31.6% | estimated ± 6.1 pp, low confidence |