CompliBench - Insurance - Violation Detection: leaderboard
Metric: Violation detection accuracy (%): share of violating assistant turns in the insurance domain that the judge flags as violations, of an LLM judge on CompliBench's synthesized multi-turn dialogues with injected, adversarially optimized guideline violations, mean of four runs at default reasoning effort; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro | 95.26 |
| 2 | Kimi K2.5 | 93.09 |
| 3 | GLM-5 | 91.85 |
| 4 | Claude Sonnet 4.6 | 88.86 |
| 5 | DeepSeek V3.2 (Thinking) | 86.26 |
| 6 | Qwen 3.5 Plus | 86.26 |
| 7 | GPT-5 | 83.93 |
| 8 | Qwen 3 Max | 78.96 |
| 9 | Qwen 3 30B A3B | 78.11 |
| 10 | Qwen 3 32B | 57.84 |
| 11 | GPT-4o | 54.89 |
| 12 | Qwen 3 14B | 43.71 |
| 13 | Qwen 3 4B | 36.88 |
| 14 | Qwen 3 8B | 31.99 |
| 15 | GPT-4o Mini | 26.94 |
Interactive version: theaggregate.ai/benchmark?slug=complibench-insurance-violation-detection · How It Works · Data refreshed daily, snapshot 2026-10-07.