Calibration
GDPval-AA → CyberGym
CyberGym is estimated from GDPval-AA with a Michaelis–Menten curve fitted on 17 models measured on both: y = 1.9858·x / (0.70257 + x), R² = 0.81, cross-validated error 6.2 pp. It is used for 31 estimates.
| Estimated model | GDPval-AA | CyberGym | Source |
|---|---|---|---|
| A.X K2 | 23.2% | 49.3% | estimated ± 6.2 pp, low confidence |
| Apodex 1.1 Mini | 34.8% | 65.8% | estimated ± 6.2 pp, medium confidence |
| DeepSeek V3 0324 | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |
| Gemma 4 12B | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |
| Gemma 4 26B A4B | 3.4% | 9.2% | estimated ± 6.2 pp, low confidence |
| Gemma 4 E2B | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |
| Gemma 4 E4B | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |
| GPT-4.1 mini | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |
| GPT-4.1 nano | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |
| GPT-4o | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |
| GPT-4o mini | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |
| GPT-5.1 | 16.5% | 37.8% | estimated ± 6.2 pp, low confidence |
| GPT-5 (high) | 21.0% | 45.7% | estimated ± 6.2 pp, low confidence |
| GPT-6 Luna | 46.9% | 79.5% | estimated ± 6.2 pp, medium confidence |
| Granite 4.2 30B | 3.9% | 10.4% | estimated ± 6.2 pp, low confidence |
| Granite 4.2 3B | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |
| Grok 4.7 | 60.8% | 92.1% | estimated ± 6.2 pp, low confidence |
| K-Exaone | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |
| LFM2.5-2.6B | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |
| Ling 2.6 Flash | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |
| Ling 3.0 Flash VL | 33.2% | 63.7% | estimated ± 6.2 pp, medium confidence |
| Ling 3.0 Tiny | 3.2% | 8.7% | estimated ± 6.2 pp, low confidence |
| Mercury 2.5 | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |
| MiMo-V2-Flash | 6.6% | 17.1% | estimated ± 6.2 pp, low confidence |
| MiniCPM5-2B | 10.5% | 25.8% | estimated ± 6.2 pp, low confidence |
| Mistral Large 4 | 46.2% | 78.8% | estimated ± 6.2 pp, medium confidence |
| Nemotron 3 Nano Omni 30B A3B | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |
| North Mini Code | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |
| Qwen3.6 Plus | 24.7% | 51.7% | estimated ± 6.2 pp, low confidence |
| Solar Pro 3 | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |
| Ultravox v0.6 Llama 3.3 70B | 0.0% | 0.0% | estimated ± 6.2 pp, low confidence |