RemoteExploitBench (N-day remote exploitation) (RemoteExploitBench)
We show this table for reference; we do not rank on it.
OpenAI's benchmark of whether a model can build a working exploit for a known vulnerability and use it against a remote target, on real applications such as GitLab, PostgreSQL and Jenkins deployed in a controlled environment.
Benchmark score on RemoteExploitBench — November 6, 2026
We compile the RemoteExploitBench rows from provider self-reports. GPT-6 Cyber leads the table at 94.7%. We do not use these results to rank models overall.
1 modelAgenticCurrentDisplay onlyUpdated November 6, 2026
Benchmark score table (1 model)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | GPT-6 CyberOpenAI | 94.7% | Not reported | Closed |
How to read this leaderboard
Operator receipt: 1 sourced row is currently displayable on this page; the leading published row is GPT-6 Cyber at 94.7%.
Honest limit: Provider-built and provider-run; task count and harness are not published, and refusals score as failures.
About RemoteExploitBench
Year
2026
Tasks
Known (N-day) vulnerabilities in deployed applications such as GitLab, PostgreSQL and Jenkins
Format
N-day target success rate
Difficulty
Remote exploit development
OpenAI introduced RemoteExploitBench with GPT-6 Cyber on November 5, 2026 and reports N-day target success: GPT-6 Cyber 94.7%, with its only failure an exploit it could not get to work remotely. GPT-6 Astra refused every task with Daybreak Blue or standard safeguards, so its 0% measures refusal rather than capability and BenchLM does not store it. OpenAI ran every result and has not released the task set, so the benchmark is display-only.
Freshness and provenance
Version
RemoteExploitBench (N-day remote exploitation)
Refresh cadence
Static
Staleness state
Current
Question availability
Private provider task set with published aggregate results
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does RemoteExploitBench measure?
OpenAI's benchmark of whether a model can build a working exploit for a known vulnerability and use it against a remote target, on real applications such as GitLab, PostgreSQL and Jenkins deployed in a controlled environment.
Which model scores highest on RemoteExploitBench?
GPT-6 Cyber by OpenAI currently leads with a score of 94.7% on RemoteExploitBench.
How many models are evaluated on RemoteExploitBench?
1 AI models have published results on RemoteExploitBench in the BenchLM catalog.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 6,100+ subscribers.
One email each week. Unsubscribe anytime.