CompliBench - Airline: leaderboard
Metric: Conversation-level accuracy (%): share of airline conversations in which the judge predicts both the governing guideline and the violation label correctly at every assistant turn, 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: Around January 2027. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro | 49.09 |
| 2 | GPT-5 | 47.26 |
| 3 | DeepSeek V3.2 (Thinking) | 42.07 |
| 4 | Qwen 3.5 Plus | 41.16 |
| 5 | GLM-5 | 31.4 |
| 6 | Kimi K2.5 | 25.91 |
| 7 | Qwen 3 Max | 24.39 |
| 8 | Qwen 3 30B A3B | 21.95 |
| 9 | Claude Sonnet 4.6 | 21.18 |
| 10 | GPT-4o | 17.07 |
| 11 | Qwen 3 32B | 8.54 |
| 12 | Qwen 3 4B | 3.66 |
| 13 | GPT-4o Mini | 2.74 |
| 14 | Qwen 3 14B | 2.13 |
| 15 | Qwen 3 8B | 1.22 |
Interactive version: theaggregate.ai/benchmark?slug=complibench-airline · How It Works · Data refreshed daily, snapshot 2026-10-07.