SIS-Bench - Action Prediction: leaderboard
Metric: Accuracy (%; 263 action prediction questions; four-option multiple choice on UAV videos (at most 32 frames), zero-shot, open models through vLLM with at most 128 generated tokens and proprietary models through their APIs; higher is better). Source: arxiv.org. Saturation forecast: Around 2029. 26 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.5 Plus | 69.2 |
| 2 | GPT-5.4 | 65.8 |
| 3 | Kimi K2.5 | 65.4 |
| 4 | Qwen 3 VL 8B (Thinking) | 64.6 |
| 5 | Seed 1.8 | 63.1 |
| 6 | Gemini 3 Flash (Preview) | 61.6 |
| 7 | Qwen 2.5 VL 7B Instruct | 60.8 |
| 8 | Qwen 3 VL 30B A3B Instruct | 59.7 |
| 9 | Qwen 2 VL 7B Instruct | 58.9 |
| 10 | Qwen 3 VL 8B Instruct | 57 |
| 11 | Qwen 3 VL 4B Instruct | 57 |
| 12 | InternVL3-14B | 55.5 |
| 13 | InternVL3.5-8B | 55.1 |
| 14 | Step3 VL 10B | 54.4 |
| 15 | GLM-4.1V-9B (Thinking) | 48.7 |
Interactive version: theaggregate.ai/benchmark?slug=sis-bench-action-prediction · How It Works · Data refreshed daily, snapshot 2026-09-29.