# ExploitGym Hardened (stronger exploit mitigations) (ExploitGym Hardened)

> A hardened variant of ExploitGym that adds address-space randomization and stack canaries to userspace targets, enables the V8 sandbox and uses stricter kernel configurations where supported.

Canonical page: https://benchlm.ai/benchmarks/exploitgymhardened

- Category: [Agentic](/agentic)
- Last updated: November 11, 2026

## About ExploitGym Hardened

- Year: 2026
- Tasks: ExploitGym tasks with stronger exploit mitigations
- Format: Intended-exploit success rate
- Difficulty: Exploitation against mitigations
- Paper: [ExploitGym: Can AI Agents Turn Security Vulnerabilities into Real Attacks?](https://arxiv.org/abs/2605.11086)

OpenAI reports the intended-exploit success rate, the same metric as standard ExploitGym: GPT-6 Cyber 36.94% and GPT-6 Astra 29.69%, against 48.45% and 42.43% on the standard suite. The mitigation modes come from the ExploitGym authors (UC Berkeley RDI): the benchmark repository ships hardened userspace builds, a V8 sandbox build and kernel defense settings, and the paper's mitigation table reports successes with defenses on for Claude Mythos Preview (45 of 898 tasks), GPT-5.5 (21) and GPT-5.4 (3), with Claude Opus 4.6, Claude Opus 4.7, Gemini 3.1 Pro and GLM-5.1 at zero. BenchLM stores paper rows as a share of the 898 tasks and OpenAI's rows as reported on its 869-task v1 implementation, so the two are not a like-for-like comparison. Display-only and separate from the ExploitGym lane.

ExploitGym Hardened is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (5 models)

| Rank | Model | Creator | Score |
|------|-------|---------|-------|
| 1 | [GPT-6 Cyber](/models/gpt-6-cyber) | OpenAI | 36.9% |
| 2 | [GPT-6 Astra](/models/gpt-6-astra) | OpenAI | 29.7% |
| 3 | [Claude Mythos Preview](/models/claude-mythos-preview) | Anthropic | 5.0% |
| 4 | [GPT-5.5](/models/gpt-5-5) | OpenAI | 2.3% |
| 5 | [GPT-5.4](/models/gpt-5-4) | OpenAI | 0.3% |

## FAQ

### What does ExploitGym Hardened measure?

A hardened variant of ExploitGym that adds address-space randomization and stack canaries to userspace targets, enables the V8 sandbox and uses stricter kernel configurations where supported.

### Which model scores highest on ExploitGym Hardened?

GPT-6 Cyber by OpenAI currently leads with a score of 36.9% on ExploitGym Hardened.

### How many models are evaluated on ExploitGym Hardened?

5 AI models have published results on ExploitGym Hardened in the BenchLM catalog.

### Does ExploitGym Hardened affect BenchLM's overall score?

Not directly. ExploitGym Hardened is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Compare Top Models on ExploitGym Hardened

- [GPT-6 Cyber vs GPT-6 Astra](/compare/gpt-6-astra-vs-gpt-6-cyber)
- [GPT-6 Astra vs Claude Mythos Preview](/compare/claude-mythos-preview-vs-gpt-6-astra)
- [Claude Mythos Preview vs GPT-5.5](/compare/claude-mythos-preview-vs-gpt-5-5)
- [GPT-5.5 vs GPT-5.4](/compare/gpt-5-4-vs-gpt-5-5)