Drive-P2D - Optimal Action (Decision-1): leaderboard
The Drive-P2D task of selecting the ego vehicle's optimal action (single-choice). Drive-P2D (Tang et al., Zhejiang University, 2026) poses human-annotated selection questions on front-camera driving images from nuScenes, KITTI and BDD100K; models reason inside think tags and only the final answer is scored. Accuracy in percent (exact match), zero-shot over all scenarios and averaged over two runs; higher is better.
Metric: Accuracy (%; select the ego vehicle's optimal 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 | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 | 56.95 | #492 |
| 2 | Qwen 2.5 VL 72B Instruct | 54.69 | #845 |
| 3 | Gemma 3 27B (IT) | 50.4 | #1306 |
| 4 | Gemini 2.5 Pro | 49.47 | #262 |
| 5 | Gemma 3 4B (IT) | 45.78 | #3848 |
| 6 | Llama 3.2 90B Vision Instruct | 45.14 | #1757 |
| 7 | Phi-4 Multimodal Instruct | 44.58 | #3914 |
| 8 | Llama 3.2 11B Instruct | 41.84 | #5039 |
| 9 | Qwen 2.5 VL 7B Instruct | 40.87 | #1842 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=drive-p2d-optimal-action-decision-1 · How It Works · Data refreshed daily, snapshot 2026-11-12.