# ExploitGym Hardened leaderboard

external benchmark

> Every figure below is reproduced as its upstream source published it: nothing is modelled, estimated, interpolated or converted. `Unavailable` means no source published the value — it is never a zero. Each value carries its evidence state and the date it was observed.

Page: https://modelchoice.dev/leaderboards/benchmarks/external-exploit-gym-hardened

ExploitGym Hardened (stronger exploit mitigations). Every model the catalog carries a published ExploitGym Hardened value for, ranked by that value.

## Benchmark record

| Field | Value |
| --- | --- |
| Category | external |
| Measure | Intended-exploit success rate |
| Tasks | ExploitGym tasks with stronger exploit mitigations |
| Difficulty | Exploitation against mitigations |
| Published by | [ExploitGym: Can AI Agents Turn Security Vulnerabilities into Real Attacks?](https://arxiv.org/abs/2605.11086) |

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.

## ExploitGym Hardened ranking

5 models with a published ExploitGym Hardened 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 | Intended-exploit success rate | Evidence |
| --- | --- | --- | --- | --- |
| 1 | [GPT-6 Cyber](https://modelchoice.dev/models/gpt-6-cyber) | OpenAI | 36.9 | Observed 2026-11-21 · source benchlm:models |
| 2 | [GPT-6 Astra](https://modelchoice.dev/models/gpt-6-astra) | OpenAI | 29.7 | Observed 2026-11-21 · source benchlm:models |
| 3 | [Claude Mythos Preview](https://modelchoice.dev/models/claude-mythos-preview) | Anthropic | 5.0 | Observed 2026-11-21 · source benchlm:models |
| 4 | [GPT-5.5](https://modelchoice.dev/models/gpt-5-5) | OpenAI | 2.3 | Observed 2026-11-21 · source benchlm:models |
| 5 | [GPT-5.4](https://modelchoice.dev/models/gpt-5-4) | OpenAI | 0.3 | Observed 2026-11-21 · source benchlm:models |

Rows are ordered by the value the source published, highest first. The source does not state whether a higher value is the better result, so this document does not either: for a benchmark that measures a rate of failure, read the table from the bottom.
