Drive-P2D - Weather and Lighting (Scene-1): leaderboard
Metric: Accuracy (%; recognize weather and illumination conditions (multiple choice, partial credit for a correct subset); zero-shot, all scenarios, mean of two runs). Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4.1 | 92.54 |
| 2 | Llama 3.2 90B Vision Instruct | 91.75 |
| 3 | Phi-4 Multimodal Instruct | 91.01 |
| 4 | Llama 3.2 11B Instruct | 87 |
| 5 | Qwen 2.5 VL 72B Instruct | 77.19 |
| 6 | Gemini 2.5 Pro | 76.73 |
| 7 | Qwen 2.5 VL 7B Instruct | 65.16 |
| 8 | Gemma 3 27B (IT) | 63.45 |
| 9 | Gemma 3 4B (IT) | 27.34 |
Interactive version: theaggregate.ai/benchmark?slug=drive-p2d-weather-and-lighting-scene-1 · How It Works · Data refreshed daily, snapshot 2026-09-26.