VideoZeroBench - Spatial Grounding: leaderboard
Metric: Mean visual IoU (%) of the bounding boxes the model predicts at the given key-frame timestamps against the annotated boxes, averaged over frames (0 for a missing box or a question without spatial evidence), on VideoZeroBench's 500 manually annotated open-ended questions (exact short answers) over 138 long videos in 13 domains; Gemini models get the raw video through the API, the others 1 fps frames up to the printed cap, temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around March 2028. 17 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Seed 2.0 Pro | 21.8 |
| 2 | GPT-5.2 | 12.6 |
| 3 | Gemini 3 Pro | 9.6 |
| 4 | Gemini 2.5 Pro | 8.6 |
| 5 | Qwen 3.5 397B A17B | 8 |
| 6 | Gemini 2.5 Flash | 4.9 |
| 7 | Qwen 3 VL 235B A22B | 3.6 |
| 8 | Qwen 2.5 VL 7B | 1.4 |
Interactive version: theaggregate.ai/benchmark?slug=videozerobench-spatial-grounding · How It Works · Data refreshed daily, snapshot 2026-10-07.