LinkS2Bench - Relative Direction: leaderboard
Metric: Relative Direction: multiple-choice accuracy (%) on the direction of a target relative to a marked satellite region 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 2030. 18 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.5 35B A3B | 38.3 |
| 2 | Gemini 3.1 Pro (Preview) | 29.9 |
| 3 | GPT-5.4 | 24.4 |
| 4 | Qwen 3.5 397B A17B | 23.9 |
| 5 | Seed 1.8 | 23.3 |
| 6 | Claude Sonnet 4.6 | 21.3 |
| 7 | Kimi K2.5 | 20.4 |
| 8 | Ministral 3 8B | 16.1 |
| 9 | Qwen 3.5 9B | 16 |
| 10 | GPT-5.4 Mini | 14 |
| 11 | Ministral 3 3B | 13.8 |
| 12 | Ministral 3 14B | 13 |
| 13 | Llama 4 Scout | 12.2 |
| 14 | Qwen 3.5 4B | 12.1 |
Interactive version: theaggregate.ai/benchmark?slug=links2bench-relative-direction · How It Works · Data refreshed daily, snapshot 2026-10-07.