AgentDrive - Hybrid: leaderboard
Evaluates text reasoning about autonomous-driving scenarios on the 2,000-example AgentDrive-MCQ evaluation. Published model/version labels are preserved. This series reports Hybrid using Accuracy (%). Source: the original paper result table.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Not forecast. 51 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | ChatGPT 4o | 72.5 | #296 |
| 2 | GPT-5 Chat | 70 | #250 |
| 3 | Gemini 2.5 Flash | 65 | #255 |
| 4 | Mistral Medium 3.1 | 60 | #418 |
| 5 | Qwen 3 235B A22B 2507 | 60 | #341 |
| 6 | GPT-4.1 Mini | 55 | #363 |
| 7 | GPT-4o (2024-11-20) | 55 | #392 |
| 8 | DeepSeek V3 (0324) | 55 | #349 |
| 9 | Llama 3.1 70B Instruct | 52.5 | #567 |
| 10 | Qwen 3 Max | 52.5 | #213 |
| 11 | Gemma 3 12B (IT) | 50 | #672 |
| 12 | GPT-4.1 | 50 | #258 |
| 13 | Kimi K2 0905 | 50 | #292 |
| 14 | Llama 3.3 70B Instruct | 47.5 | #541 |
| 15 | Phi-4-reasoning-plus | 45 | #1339 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=agentdrive-hybrid · How It Works · Data refreshed daily, snapshot 2026-11-05.