ExploitGymBrowse 284 families
ExploitGym
Can the model diagnose or complete difficult cybersecurity tasks?
Version / measure
- Domain
- Cybersecurity
- Ranked models
- 3
- Reported values
- 3
- Reports
- 1 · lab
- Direction
- Higher is better
- Status
- Latest stable
Rankings
Best reported score per model across effort settings
| # | Model | Best setting | Score |
|---|---|---|---|
| 1 | GPT-5.6 SolOpenAI | maxcost n/a1 value · Sep 3 · source | 30.3% |
| 2 | GPT-6 SolOpenAI | reportedcost n/a1 value · Sep 22 · source | 22.1% |
| 3 | GPT-6 LunaOpenAI | reportedcost n/a1 value · Sep 22 · source | 11.6% |
Effort curve
Score vs. reported cost. Lines connect complete sweeps only.
No cost-linked effort sweep for this version.
Definition and comparison boundarylimited methodology
Can the model diagnose or complete difficult cybersecurity tasks?
Security analysis or exploitation under a controlled evaluation environment. Higher is better. The value is the percentage reported in this lab's table.
Harness, tools, prompts, attempt count, and benchmark version can materially change the result. Compare within one reporting context.
OpenAI reports GPT-5.6 Sol at Max 30.3%, GPT-6 Sol 22.1%, and GPT-6 Luna 11.6%. Harness details and effort for the GPT-6 variants are not specified in the card.
Method / source ↗