ExploitBench AutoNudge Flag Capture Rate (Sonnet 5 & Opus 5 System Cards): leaderboard
ExploitBench (Lee and Brumley, 2026) on 41 V8 vulnerabilities with 16 capability flags. Cap% is the percentage of available flags captured per environment over a random subset of three AutoNudge-arm trials, averaged across environments (higher is better). Anthropic's runs from the Claude Opus 5 (p. 37) and Sonnet 5 (p. 31) system cards.
Metric: Flags Captured per Environment (%). Source: www-cdn.anthropic.com. Saturation forecast: Around February 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Mythos 5 | 78 | #8 |
| 2 | Claude Opus 5 | 70 | #16 |
| 3 | Claude Opus 4.8 | 40 | #39 |
| 4 | Claude Sonnet 5 | 31 | #65 |
| 5 | Claude Sonnet 4.6 | 24 | #94 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=exploitbench-autonudge-flag-capture-rate-sonnet-5-opus-5-system-cards · How It Works · Data refreshed daily, snapshot 2026-11-05.