DashBench - Q4 Grounded Reasoning: leaderboard
DashBench evaluates video understanding on 1,709 driving clips. Five explanation prompts and binary event detection are scored separately; mean agreement excludes the binary task. The 823-clip corner-case-window subset and training-free decoder variants retain their own conditions. Q5 asks for definition-guided corner-case identification and description. This board reports lLM-judged answer accuracy (%). Higher is better.
Metric: LLM-judged answer accuracy (%). Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked. Updated 14 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-5 | 46.52 | #167 | |
| 2 | Gemini 3 Flash | 39.58 | #163 | |
| 3 | Qwen 3 VL 235B A22B | 25.92 | #484 | |
| 4 | Qwen 3 VL 8B | 22 | #899 | |
| 5 | InternVL3.5-241B | 19.13 | #625 | |
| 6 | Qwen 3 VL 32B | 18.78 | #609 | |
| 7 | Ovis2.5-9B | 18.72 | #1236 | |
| 8 | MiniCPM-V-4.5 | 17.27 | #1390 | |
| 9 | InternVL3.5-38B | 16.44 | #986 | |
| 10 | InternVL3.5-8B | Evaluation run: DashBench SECOND | 13.4 | #1558 |
| 11 | InternVideo2.5-Chat-8B | 9.95 | #2017 | |
| 12 | InternLM-XComposer2.5 | 9.07 | #3981 | |
| 13 | LLaVA-OV-1.5-8B | 7.14 | #4771 | |
| 14 | GLM-4.1V-9B (Thinking) | 6.32 | ||
| 15 | VITA-1.5 | 6.03 | #3196 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=dashbench-q4-grounded-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-16.