LinkS2Bench - Relative Distance: leaderboard

LinkS2Bench (LinkS²Bench, 2026) task on drone videos paired with satellite imagery: multiple-choice accuracy (%) on how a target's distance to a marked satellite region changes; higher is better.

Metric: Relative Distance: multiple-choice accuracy (%) on how a target's distance to a marked satellite region changes 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. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 3.5 397B A17B37.5#270
2GPT-5.4 Mini33.3#409
3Qwen 3.5 35B A3B31.9#494
4Gemini 3.1 Pro (Preview)30.6#102
5GPT-5.429.8#138
6Seed 1.829#258
7LLaVA-OneVision-1.5-8B27.2#1876
8Gemini-3.1-Flash [LinkS2Bench checkpoint unspecified]25
9Qwen 3.5 4B23.8#1270
10GLM-4.6V-Flash23.5#1049
11Qwen 3.5 9B22.7#838
12Claude Sonnet 4.619#179
13Ministral 3 3B18.5#3606
14LLaVA-OneVision-1.5-4B18.5#1874
15Kimi K2.517.4#272

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=links2bench-relative-distance · How It Works · Data refreshed daily, snapshot 2026-11-22.