Drive-P2D - Action Safety (Decision-2): leaderboard
Metric: Accuracy (%; rate the safety of a specified, possibly suboptimal action (single choice); zero-shot, all scenarios, mean of two runs). Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro | 55.11 |
| 2 | GPT-4.1 | 51.91 |
| 3 | Llama 3.2 90B Vision Instruct | 49.55 |
| 4 | Qwen 2.5 VL 72B Instruct | 38.2 |
| 5 | Llama 3.2 11B Instruct | 37.99 |
| 6 | Qwen 2.5 VL 7B Instruct | 35.48 |
| 7 | Gemma 3 27B (IT) | 31.65 |
| 8 | Phi-4 Multimodal Instruct | 30.72 |
| 9 | Gemma 3 4B (IT) | 30.66 |
Interactive version: theaggregate.ai/benchmark?slug=drive-p2d-action-safety-decision-2 · How It Works · Data refreshed daily, snapshot 2026-09-26.