safety benchmark
ExploitGym Leaderboard
ExploitGym is a benchmark based on real-world vulnerabilities in userspace programs, Google's V8 engine, and the Linux kernel, in which agents craft end-to-end exploits for unauthorized code execution.
Updated Sep 5, 2026
Models5
Model coverage5
MetricScore
EvidenceB
ExploitGym Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelOP | Score42.40% | Percentile100.00% | Participants5 | EvidenceC | Evaluated |
| Rank02 | ModelOP | Score33.70% | Percentile75.00% | Participants5 | EvidenceC | Evaluated |
| Rank03 | ModelOP | Score23.20% | Percentile50.00% | Participants5 | EvidenceC | Evaluated |
| Rank04 | ModelZA | Score15.00% | Percentile25.00% | Participants5 | EvidenceC | Evaluated |
| Rank05 | ModelOP | Score12.40% | Percentile0.00% | Participants5 | EvidenceC | Evaluated |
ExploitGym Highlights
The leading models and scores on this benchmark.
ExploitGym Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for ExploitGym
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What is ExploitGym?
What ExploitGym measures and how its scores work.
ExploitGym is a safety benchmark that provides a vulnerability and a proof-of-vulnerability input and requires agents to create a working end-to-end exploit.
It measures performance using Score, reported as a ratio, for achieving unauthorized code execution.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- ExploitGym
- Modality
- text
- Primary category
- safety
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- overall
LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about ExploitGym.
Which model scores highest on ExploitGym?
GPT-6 Astra is currently ranked first with 42.40%.
What are the top three models on ExploitGym?
The current leaders are GPT-6 Astra (42.40%), GPT-5.6 Sol (33.70%), and GPT-5.6 Terra (23.20%).
Which ExploitGym model has the lowest official input price?
GPT-5.6 Luna has the lowest matched official input price at $0.20 input / $1.2 output per 1M tokens.
Which models are fastest among ExploitGym results?
The fastest matched records are GLM-5.3 (472.69 tok/s via FriendliAI), GPT-5.6 Terra (99.35 tok/s via OpenAI), and GPT-5.6 Luna (41.87 tok/s via OpenAI).
Does the highest score result prove overall model quality?
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.