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

#ModelSettingScoreOverall rank
1GPT-546.52#167
2Gemini 3 Flash39.58#163
3Qwen 3 VL 235B A22B25.92#484
4Qwen 3 VL 8B22#899
5InternVL3.5-241B19.13#625
6Qwen 3 VL 32B18.78#609
7Ovis2.5-9B18.72#1236
8MiniCPM-V-4.517.27#1390
9InternVL3.5-38B16.44#986
10InternVL3.5-8BEvaluation run: DashBench SECOND13.4#1558
11InternVideo2.5-Chat-8B9.95#2017
12InternLM-XComposer2.59.07#3981
13LLaVA-OV-1.5-8B7.14#4771
14GLM-4.1V-9B (Thinking)6.32
15VITA-1.56.03#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.