# AA AutomationBench → ExploitGym

Calibration · https://benchgap.net/calibration/aa-automation-bench/current/to/exploitgym/current · data retrieved 2026-11-18

ExploitGym is estimated from AA AutomationBench with a Hill curve fitted on 7 models measured on both: y = 0.0561 + (1.2000 − 0.0561)·x^6.00 / (0.87948^6.00 + x^6.00), R² = 0.74, cross-validated error 9.8 pp. It is used for 3 estimates.

| Estimated model | AA AutomationBench | ExploitGym | Source |
|---|---|---|---|
| [GPT-6.2 Astra](https://benchgap.net/model/gpt-6-2-astra.md) | 77.2% | 41.5% | estimated ± 9.8 pp, low confidence |
| [Qwen4 Max](https://benchgap.net/model/qwen4-max.md) | 72.6% | 33.1% | estimated ± 9.8 pp, low confidence |
| [GPT-6.2 Sol](https://benchgap.net/model/gpt-6-2-sol.md) | 71.3% | 30.9% | estimated ± 9.8 pp, low confidence |

## Fitted on

| Model | AA AutomationBench | ExploitGym |
|---|---|---|
| [Claude Opus 5.5](https://benchgap.net/model/claude-opus-5-5.md) | 69.5% | 33.3% |
| [Gemini 4 Argon](https://benchgap.net/model/gemini-4-argon.md) | 77.5% | 39.5% |
| [Gemini 4 Carbon](https://benchgap.net/model/gemini-4-carbon.md) | 77.8% | 53.5% |
| [GLM-5.3](https://benchgap.net/model/glm-5-3.md) | 62.2% | 15.0% |
| [GPT-6 Luna](https://benchgap.net/model/gpt-6-luna.md) | 53.2% | 11.6% |
| [Kimi K3.1 Pro](https://benchgap.net/model/kimi-3-1-pro.md) | 73.3% | 20.7% |
| [MiMo-V2.6-Pro](https://benchgap.net/model/mimo-v2-6-pro.md) | 58.6% | 17.8% |
