Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Agentic benchmark
Artificial Analysis Cyber Index. Every model the catalog carries a published AA Cyber Index value for, ranked by that value.
Artificial Analysis' composite index of cyber-defense capability: finding vulnerabilities in real codebases, reproducing and validating them, and patching them without breaking existing functionality. It averages CWE-Bench-AA, DeepsecBench-AA and CyberGym-E2E-AA. 21 models in the ModelScale catalog have a published value.
Each point is a model with a published AA Cyber Index value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
21 models with a published AA Cyber Index value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Index score (0-100) |
|---|---|---|---|
| 1 | Moonshot AI | 64.1 | |
| 2 | xAI | 56.4 | |
| 3 | Xiaomi | 56.1 |
What the benchmark measures and how its value is published, from the source’s own record.
AA Cyber Index (Artificial Analysis Cyber Index) is an agentic benchmark published by Artificial Analysis Cyber Index. Artificial Analysis' composite index of cyber-defense capability: finding vulnerabilities in real codebases, reproducing and validating them, and patching them without breaking existing functionality. It averages CWE-Bench-AA, DeepsecBench-AA and CyberGym-E2E-AA. Its evaluation set is described by the source as three defensive cybersecurity evaluations. The source rates its difficulty as real-world vulnerability discovery and patching.
Each model's result is published by Artificial Analysis Cyber Index as index score (0-100). ModelScale prints that value to 1 decimal place, exactly as published, and never converts or rescales it. This page ranks the 21 models that carry a published value, highest value first; a model the source never scored is left out rather than ranked at zero. The source does not state whether a higher value is the better result, so neither does this page.
| 4 | OpenAI | 52.7 |
| 5 | Mistral | 50.9 |
| 6 | Z.AI | 49.9 |
| 7 | StepFun | 47.7 |
| 8 | Moonshot AI | 41.3 |
| 9 | DeepSeek | 40.8 |
| 10 | OpenAI | 37.7 |
| 11 | OpenAI | 37.3 |
| 12 | OpenAI | 36.8 |
| 13 | Z.AI | 36.4 |
| 14 | RABeam 1.0 501B-A23B | Reflection AI | 33.7 |
| 15 | OpenAI | 33.4 |
| 16 | Anthropic | 28.7 |
| 17 | 25.1 |
| 18 | Anthropic | 25.0 |
| 19 | MiniMax | 22.4 |
| 20 | NVIDIA | 13.1 |
| 21 | Alibaba | 12.7 |
Evidence key: Observed
Rows are ordered by the value Artificial Analysis Cyber Index published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.