MMVIAD: leaderboard
Metric: Mean of three accuracies and visible-time mIoU (%; MMVIAD-Standard test split, 1,101 clips, 4,404 questions, 32 frames). Source: arxiv.org. Saturation forecast: Around October 2027. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 61.8 |
| 2 | GPT-5.4 | 56.4 |
| 3 | GPT-5.4 Mini | 52.9 |
| 4 | Qwen 3 VL 8B | 41.6 |
Interactive version: theaggregate.ai/benchmark?slug=mmviad · How It Works · Data refreshed daily, snapshot 2026-09-26.