Autonomous Pentesting on AutoPenBench
5Attack Success CountQwen3-32B-finetune (xOffense)
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Qwen3-32B-finetune (xOffense)Backbone=Qwen3-32B, Status=Fine-tuned2025.09 | 5 | 5 | 5 | 3 | 6 | 24 | |
| Llama3.1-405B (VulnBot)Backbone=Llama3.1-405B, Framework=VulnBot2025.09 | 3 | 2 | 2 | 0 | 3 | 10 | |
| Qwen3-32B (Base)Backbone=Qwen3-32B, Status=Base2025.09 | 2 | 2 | 3 | 0 | 3 | 10 | |
| GPT-4oBackbone=GPT-4o2025.09 | 1 | 2 | 3 | 0 | 1 | 7 | |
| Llama3.3-70B (VulnBot)Backbone=Llama3.3-70B, Framework=VulnBot2025.09 | 1 | 1 | 2 | 0 | 2 | 6 | |
| Llama3.1-405B (PentestGPT)Backbone=Llama3.1-405B, Framework=PentestGPT2025.09 | 1 | 0 | 2 | 0 | 0 | 3 |