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

#ModelSettingScoreOverall rank
1GPT-536.92#167
2Gemini 3 Flash35.28#163
3Qwen 3 VL 235B A22B20.19#484
4Qwen 3 VL 32B18.9#609
5Qwen 3 VL 8B15.51#899
6Ovis2.5-9B15.41#1236
7LlaVA-OV-RS (7B)14.86
8InternVL3.5-38B13.22#986
9Tarsier2-7B13.11#844
10InternVL3.5-241B12.81#625
11MiniCPM-V-4.511.18#1390
12InternLM-XComposer2.59.89#3981
13InternVideo2.5-Chat-8B9.71#2017
14InternVL3.5-8BEvaluation run: DashBench TRiCD8.72#1558
15GLM-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.