Skip to content

Benchmark

ExploitBench · June–August 2026

Scores are percentages. Higher is better.

Models measured
2
Scores recorded
3
Sources
1
IQ weight
No weight published

Score distribution

18.7%
Mean
11.5%
Median
14.6%
Std dev
5.5%
10th percentile
39%
90th percentile
5.5%
Min
5.5%39%
The band covers the middle 80% of scores. The line marks the median.

Leaderboard

#ModelScore
1GPT-6 AstraJune-August 2026 · 300-turn limit· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.39%
2GPT-5.6 Solsimilar settings · fewer 300-turn-limit hits· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.11.5%
3GPT-5.6 SolJune-August 2026 · 300-turn limit· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.5.5%

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.