OmniTraffic - Temporal Reasoning: leaderboard
Metric: Accuracy (%) on temporal reasoning (order frames and track queues over time), 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 2030. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Grok 4 | 56.4 |
| 2 | Claude Sonnet 4.5 (Thinking) | 54.9 |
| 3 | Qwen 3 VL 235B A22B (Thinking) | 50.5 |
| 4 | Gemini 3 Pro | 48.9 |
| 5 | GPT-4o | 48.7 |
| 6 | GPT-5.2 | 43 |
| 7 | Gemini 2.5 Pro | 22.8 |
Interactive version: theaggregate.ai/benchmark?slug=omnitraffic-temporal-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-29.