# DeepsecBench-AA → HLE w/ tools

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

HLE w/ tools is estimated from DeepsecBench-AA with a offset logistic curve fitted on 8 models measured on both: y = 0.5116 + (0.6461 − 0.5116) / (1 + exp(−74.98·(x − 0.2104))), R² = 0.67, cross-validated error 5.1 pp. It is used for 4 estimates.

| Estimated model | DeepsecBench-AA | HLE w/ tools | Source |
|---|---|---|---|
| [GPT-6.2 Sol](https://benchgap.net/model/gpt-6-2-sol.md) | 51.6% | 64.6% | estimated ± 5.1 pp, low confidence |
| [GPT-6.2 Astra](https://benchgap.net/model/gpt-6-2-astra.md) | 46.8% | 64.6% | estimated ± 5.1 pp, low confidence |
| [Mistral Large 4](https://benchgap.net/model/mistral-large-4.md) | 15.0% | 51.3% | estimated ± 5.1 pp, medium confidence |
| [Qwen3.8-27B](https://benchgap.net/model/qwen3-8-27b.md) | 11.5% | 51.2% | estimated ± 5.1 pp, medium confidence |

## Fitted on

| Model | DeepsecBench-AA | HLE w/ tools |
|---|---|---|
| [Claude Opus 5](https://benchgap.net/model/claude-opus-5.md) | 25.9% | 64.7% |
| [Claude Opus 5.5](https://benchgap.net/model/claude-opus-5-5.md) | 26.9% | 67.7% |
| [DeepSeek V4.1 Flash 1017](https://benchgap.net/model/deepseek-v4-1-flash.md) | 31.2% | 63.9% |
| [GLM-5.3](https://benchgap.net/model/glm-5-3.md) | 24.3% | 62.5% |
| [GLM-5.3-Flash](https://benchgap.net/model/glm-5-3-flash.md) | 19.9% | 55.3% |
| [GPT-6 Astra](https://benchgap.net/model/gpt-6-astra.md) | 36.9% | 57.2% |
| [Kimi K3.1 Pro](https://benchgap.net/model/kimi-3-1-pro.md) | 36.7% | 70.0% |
| [Beam 1.0 501B-A23B](https://benchgap.net/model/reflection-beam.md) | 6.9% | 51.1% |
