DriveCombo - L1 Atomic Rules: leaderboard
Metric: Accuracy (%) on DriveCombo Level 1 (one atomic traffic rule) four-option multiple-choice questions on the 20% test split, rules from five countries' traffic codes; input is the question with four RGB frames of the simulated driving scene; zero-shot, greedy decoding; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Pro | 86.91 | #75 |
| 2 | Gemini 2.5 Pro | 85.71 | #145 |
| 3 | Claude Sonnet 4.5 | 83.8 | #138 |
| 4 | GPT-5 Mini | 80.62 | #176 |
| 5 | GLM-4.5V | 80.44 | #339 |
| 6 | Gemini 2.5 Flash | 79.31 | #237 |
| 7 | GPT-5 Nano | 74.04 | #415 |
| 8 | Gemma 3 27B | 73.94 | #596 |
| 9 | Gemma 3 12B | 64.32 | #666 |
| 10 | Gemma 3 4B | 58.79 | #1084 |
| 11 | Llama 3.2 11B | 52.46 | #1183 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=drivecombo-l1-atomic-rules · How It Works · Data refreshed daily, snapshot 2026-10-11.