DashBench - Q2: 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. 22 models tracked. Updated 14 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-5 | 36.92 | #167 | |
| 2 | Gemini 3 Flash | 35.28 | #163 | |
| 3 | Qwen 3 VL 235B A22B | 20.19 | #484 | |
| 4 | Qwen 3 VL 32B | 18.9 | #609 | |
| 5 | Qwen 3 VL 8B | 15.51 | #899 | |
| 6 | Ovis2.5-9B | 15.41 | #1236 | |
| 7 | LlaVA-OV-RS (7B) | 14.86 | ||
| 8 | InternVL3.5-38B | 13.22 | #986 | |
| 9 | Tarsier2-7B | 13.11 | #844 | |
| 10 | InternVL3.5-241B | 12.81 | #625 | |
| 11 | MiniCPM-V-4.5 | 11.18 | #1390 | |
| 12 | InternLM-XComposer2.5 | 9.89 | #3981 | |
| 13 | InternVideo2.5-Chat-8B | 9.71 | #2017 | |
| 14 | InternVL3.5-8B | Evaluation run: DashBench TRiCD | 8.72 | #1558 |
| 15 | GLM-4.1V-9B (Thinking) | 5.5 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=dashbench-q2 · How It Works · Data refreshed daily, snapshot 2026-11-16.