ExploitBench Internal Port (June--August 2026) leaderboard
As of November 22, 2026, among the results listed here, OpenAI GPT-6.2 Astra has the highest reported ExploitBench Internal Port (June--August 2026) score, 53.5% (Mean arbitrary code execution rate across V8 and Chrome), published by OpenAI, ahead of OpenAI GPT-6.2 Sol at 50.5%. 5 models have a ExploitBench Internal Port (June--August 2026) result in our records.
Scores as the model makers published them, not re-run by us; each row links to its announcement.
Best reported score
53.5%
Reported by OpenAI, November 17, 2026 · Mean arbitrary code execution rate across V8 and Chrome
Runner-up
50.5%
Reported by OpenAI, November 17, 2026 · Mean arbitrary code execution rate across V8 and Chrome
5 models ranked · 1 maker
| Rank | Model | Score | Setting | Reported by | Date |
|---|---|---|---|---|---|
| 1 | GPT-6.2 Astra | 53.5% | Mean arbitrary code execution rate across V8 and Chrome | OpenAI | November 17, 2026 (model release) |
| 2 | GPT-6.2 Sol | 50.5% | Mean arbitrary code execution rate across V8 and Chrome | OpenAI | November 17, 2026 (model release) |
| 3 | GPT-6 Astra | 31.5% | Mean arbitrary code execution rate across V8 and Chrome | OpenAI | November 17, 2026 (model release) |
| 4 | GPT-6.1 Terra | 24.5% | Mean arbitrary code execution rate across V8 and Chrome | OpenAI | November 17, 2026 (model release) |
| 5 | GPT-6.1 Sol | 21.5% | Mean arbitrary code execution rate across V8 and Chrome | OpenAI | November 17, 2026 (model release) |
How much to trust these numbers
- Source
- Reported by the model makers themselves. Not re-run by LLMBenchmark or anyone independent.
- Settings
- All ranked results use the same reported setting.
- Spread
- 53.5% to 21.5% across 5 models.
- Freshness
- Published November 17, 2026.
What ExploitBench Internal Port (June--August 2026) measures
It asks models to develop working exploits for vulnerabilities disclosed in June–August 2026; the score is the percentage of attempts achieving arbitrary code execution. Source: ExploitBench