OI-Bench - Bonus Framing - Attack Success Rate: leaderboard
Metric: Attack Success Rate (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 | 3.76 |
| 2 | Claude Haiku 4.5 | 4.6 |
| 3 | Grok 4.1 | 5.99 |
| 4 | GPT-5 Mini | 6.31 |
| 5 | Llama 4 Scout | 6.4 |
| 6 | Llama 4 Maverick | 7.15 |
| 7 | Qwen 3 8B | 11.86 |
| 8 | Qwen 3 235B A22B | 13.67 |
| 9 | DeepSeek V3.2 | 16.96 |
| 10 | DeepSeek R1 | 18.5 |
| 11 | Gemini 2.5 Pro | 26.28 |
| 12 | Gemini 2.5 Flash Lite | 31.36 |
Interactive version: theaggregate.ai/benchmark?slug=oi-bench-bonus-framing-attack-success-rate · How It Works · Data refreshed daily, snapshot 2026-09-25.