Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
external benchmark
ExploitBench — Internal Port (June–August 2026). Every model the catalog carries a published ExploitBench Internal Port value for, ranked by that value.
OpenAI's internal port of ExploitBench restricted to V8 and Chrome vulnerabilities disclosed in June to August 2026, to limit contamination from historical exploits. 9 models in the ModelChoice catalog have a published value.
Each point is a model with a published ExploitBench Internal Port value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
9 models with a published ExploitBench Internal Port 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 | Mean arbitrary-code-execution success rate |
|---|---|---|---|
| 1 | OpenAI | 53.5 | |
| 2 | OpenAI | 50.5 | |
| 3 | OpenAI | 42.0 |
What the benchmark measures and how its value is published, from the source’s own record.
ExploitBench Internal Port (ExploitBench — Internal Port (June–August 2026)) is an external benchmark published by GPT-6 Cyber System Card. OpenAI's internal port of ExploitBench restricted to V8 and Chrome vulnerabilities disclosed in June to August 2026, to limit contamination from historical exploits. Its evaluation set is described by the source as recently disclosed V8 and Chrome vulnerabilities (June–August 2026). The source rates its difficulty as browser exploitation.
Each model's result is published by GPT-6 Cyber System Card as mean arbitrary-code-execution success rate. ModelChoice prints that value to 1 decimal place, exactly as published, and never converts or rescales it. This page ranks the 9 models that carry a published value, highest value first; a model the source never scored is left out rather than ranked at zero. The source does not state whether a higher value is the better result, so neither does this page.
| 4 | OpenAI | 39.0 |
| 5 | OpenAI | 24.5 |
| 6 | OpenAI | 21.5 |
| 7 | OpenAI | 5.5 |
| 7 | OpenAI | 5.5 |
| 9 | OpenAI | 0.0 |
Evidence key: Observed
Rows are ordered by the value GPT-6 Cyber System Card published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.