# Hermes Bench → CWE-Bench-AA

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

CWE-Bench-AA is estimated from Hermes Bench with a Michaelis–Menten curve fitted on 10 models measured on both: y = 2.0000·x / (1.73862 + x), R² = 0.36, cross-validated error 4.6 pp. It is used for 2 estimates.

| Estimated model | Hermes Bench | CWE-Bench-AA | Source |
|---|---|---|---|
| [Hy4 preview](https://benchgap.net/model/hy4-preview.md) | 70.8% | 57.9% | estimated ± 4.6 pp, medium confidence |
| [Ling 3.0 Flash](https://benchgap.net/model/ling-3-0-flash.md) | 53.8% | 47.3% | estimated ± 4.6 pp, low confidence |

## Fitted on

| Model | Hermes Bench | CWE-Bench-AA |
|---|---|---|
| [Claude Opus 5.5](https://benchgap.net/model/claude-opus-5-5.md) | 76.7% | 58.3% |
| [Gemini 3.8 Flash](https://benchgap.net/model/gemini-3-8-flash.md) | 62.2% | 52.5% |
| [GLM-5.3](https://benchgap.net/model/glm-5-3.md) | 71.8% | 55.8% |
| [GLM-5.3-Flash](https://benchgap.net/model/glm-5-3-flash.md) | 68.9% | 55.8% |
| [GPT-6 Astra](https://benchgap.net/model/gpt-6-astra.md) | 73.7% | 63.3% |
| [GPT-6 Luna](https://benchgap.net/model/gpt-6-luna.md) | 70.5% | 56.7% |
| [GPT-6 Sol](https://benchgap.net/model/gpt-6-sol.md) | 73.5% | 64.2% |
| [Grok 4.7](https://benchgap.net/model/grok-4-7.md) | 74.4% | 68.3% |
| [Kimi K3](https://benchgap.net/model/kimi-k3.md) | 71.5% | 53.3% |
| [Step 5 Preview](https://benchgap.net/model/step-5-preview.md) | 69.2% | 52.5% |
