OI-Bench - Instructional Interference - 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 | 3.44 |
| 2 | Claude Haiku 4.5 | 6.2 |
| 3 | Llama 4 Maverick | 6.3 |
| 4 | Llama 4 Scout | 7.55 |
| 5 | Grok 4.1 | 8.55 |
| 6 | GPT-5 Mini | 9.35 |
| 7 | Gemini 2.5 Pro | 13.33 |
| 8 | Qwen 3 8B | 18.19 |
| 9 | Qwen 3 235B A22B | 18.82 |
| 10 | DeepSeek V3.2 | 22.13 |
| 11 | Gemini 2.5 Flash Lite | 24.07 |
| 12 | DeepSeek R1 | 24.88 |
Interactive version: theaggregate.ai/benchmark?slug=oi-bench-instructional-interference-attack-success-rate · How It Works · Data refreshed daily, snapshot 2026-09-25.