VideoZeroBench - Level-4 (Answer with Temporal Grounding): leaderboard
Metric: Level-4 accuracy (%), the share of all 500 questions answered correctly at Level-3 whose predicted evidence intervals reach temporal IoU above 0.3, 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 2029. 17 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro | 8 |
| 2 | Gemini 2.5 Pro | 7.2 |
| 3 | Gemini 2.5 Flash | 5.4 |
| 4 | Seed 2.0 Pro | 3.8 |
| 5 | Qwen 3 VL 235B A22B | 3.4 |
| 6 | GPT-5.2 | 1.8 |
| 7 | Qwen 3.5 397B A17B | 0 |
| 8 | Qwen 2.5 VL 7B | 0 |
Interactive version: theaggregate.ai/benchmark?slug=videozerobench-level-4-answer-with-temporal-grounding · How It Works · Data refreshed daily, snapshot 2026-10-07.