OmniTraffic - Scene Attributes: leaderboard

Metric: Accuracy (%) on scene attributes (weather and pedestrians in real footage), multiple-choice VQA on the 3,204-item human-verified OmniTraffic test set (12 reconstructed simulated intersections plus real roadside surveillance from South Korea and Tianjin), greedy decoding (at most 500 tokens), exact match against the answer letter with up to three attempts for invalid output (still invalid counts as wrong); higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 11 models tracked.

Top models

#ModelScore
1Gemini 2.5 Pro92.5
2Gemini 3 Pro87.5
3Qwen 3 VL 235B A22B (Thinking)87.5
4GPT-4o82.5
5GPT-5.280
6Grok 480
7Claude Sonnet 4.5 (Thinking)65

Interactive version: theaggregate.ai/benchmark?slug=omnitraffic-scene-attributes · How It Works · Data refreshed daily, snapshot 2026-09-29.