LinkS2Bench - Spatiotemporal Ordering: leaderboard
Metric: Spatiotemporal Ordering: multiple-choice accuracy (%) on the order in which marked satellite regions appear in the video on the full LinkS2Bench (17,903 questions over 1,500 UAV videos paired with satellite images), temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 18 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 44 |
| 2 | Seed 1.8 | 42.7 |
| 3 | Qwen 3.5 35B A3B | 34.7 |
| 4 | Llama 4 Scout | 31.3 |
| 5 | Qwen 3.5 397B A17B | 26.1 |
| 6 | Kimi K2.5 | 25 |
| 7 | GPT-5.4 Mini | 21.7 |
| 8 | Qwen 3.5 9B | 17.3 |
| 9 | GPT-5.4 | 16.4 |
| 10 | Ministral 3 3B | 14.9 |
| 11 | Claude Sonnet 4.6 | 13.7 |
| 12 | Ministral 3 14B | 12.5 |
| 13 | Ministral 3 8B | 12.5 |
| 14 | Qwen 3.5 4B | 6.5 |
Interactive version: theaggregate.ai/benchmark?slug=links2bench-spatiotemporal-ordering · How It Works · Data refreshed daily, snapshot 2026-10-07.