NextMotionQA - Question Answering: leaderboard
NextMotionQA (2026) tests how well vision-language models understand human motion shown as rendered 3D motion clips, with expert-verified items stratified by body part, direction, action and difficulty. This board is multi-select question answering over 511 items, scored by the Jaccard overlap in percent between predicted and gold option sets; higher is better.
Metric: Task 1 Jaccard (%) between predicted and gold option subsets on 511 multi-select four-option questions about rendered AMASS human-motion clips given to the model as RGB video (the paper headline metric for Task 1); higher is better. Source: arxiv.org. Saturation forecast: Around July 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.6 Plus | 64.52 | #222 |
| 2 | Qwen 3.5 27B | 55.78 | #359 |
| 3 | GPT-5.4 Mini | 52.27 | #389 |
| 4 | Qwen 3.5 9B | 47.95 | #803 |
| 5 | Qwen 3.5 4B | 44.19 | #1175 |
| 6 | Qwen 3.5 0.8B | 34.24 | #8081 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=nextmotionqa-question-answering · How It Works · Data refreshed daily, snapshot 2026-11-13.