LinkS2Bench - Region Reachability: leaderboard
Metric: Region Reachability: yes/no accuracy (%) on whether a target can reach a marked region under a given future action 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 June 2028. 18 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.5 397B A17B | 66.7 |
| 2 | Seed 1.8 | 63.7 |
| 3 | Llama 4 Scout | 60.4 |
| 4 | Qwen 3.5 35B A3B | 58.3 |
| 5 | Gemini 3.1 Pro (Preview) | 58.1 |
| 6 | GPT-5.4 Mini | 58 |
| 7 | GPT-5.4 | 57.6 |
| 8 | Claude Sonnet 4.6 | 48 |
| 9 | Kimi K2.5 | 40 |
| 10 | Ministral 3 8B | 38.9 |
| 11 | Ministral 3 3B | 37.8 |
| 12 | Qwen 3.5 9B | 37.3 |
| 13 | Qwen 3.5 4B | 36 |
| 14 | Ministral 3 14B | 35.8 |
Interactive version: theaggregate.ai/benchmark?slug=links2bench-region-reachability · How It Works · Data refreshed daily, snapshot 2026-10-07.