Agentic benchmark
ExploitGym leaderboard
Every model the catalog carries a published ExploitGym value for, ranked by that value.
CategoryAgentic
MeasureWorking exploit generation
Tasks898 exploitation tasks
DifficultyAdvanced cybersecurity exploitation
A controlled benchmark for evaluating whether AI agents can extend vulnerability-triggering inputs into working exploits.
ExploitGym ranking
10 models with a published ExploitGym 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 | Working exploit generation |
|---|---|---|---|
| 1 |  GPT-6 Astra | OpenAI | 42.4 |
| 2 | GPT-5.6 Sol | OpenAI | 33.7 |
| 3 | GPT-5.6 Terra | OpenAI | 23.2 |
| 4 |  Claude Mythos Preview | Anthropic | 17.5 |
| 5 |  DeepSeek V4.1 Flash | DeepSeek | 15.3 |
| 6 |  GLM-5.3 | Z.AI | 15.0 |
| 7 | GPT-5.5 | OpenAI | 13.4 |
| 8 | GPT-5.6 Luna | OpenAI | 12.4 |
| 9 | GPT-5.4 | OpenAI | 6.0 |
| 10 |  Muse Spark 1.1 | Meta | 0.8 |
Evidence key: Observed
Rows are ordered by the value ExploitGym: Can AI Agents Turn Security Vulnerabilities into Real Attacks? 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.