LinkS2Bench - Temporal Grounding: leaderboard

Metric: Temporal Grounding: ACC@1s (%): share of timestamp answers within one second of when a marked satellite region appears 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 2028. 18 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview)51.9
2Kimi K2.548
3GPT-5.444.6
4Seed 1.842.3
5Qwen 3.5 397B A17B39.5
6Qwen 3.5 9B36.8
7Qwen 3.5 35B A3B36.7
8Qwen 3.5 4B32.1
9GPT-5.4 Mini28.7
10Ministral 3 8B27.4
11Claude Sonnet 4.627.3
12Ministral 3 14B26.3
13Llama 4 Scout20.4
14Ministral 3 3B20.2

Interactive version: theaggregate.ai/benchmark?slug=links2bench-temporal-grounding · How It Works · Data refreshed daily, snapshot 2026-10-07.