CVSBench - CVUSA Satellite-to-Ground: leaderboard
Metric: Accuracy (%) on CVUSA-subset satellite-to-ground questions: from a satellite image, infer spatial attributes of the unseen street view; four-choice and binary-choice questions over paired street-view and satellite images (Gemini-2.5-flash-generated, human-verified), test split; higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3 VL 4B Instruct | 62.4 |
| 2 | GPT-4o | 62.1 |
| 3 | Qwen 3 VL 8B Instruct | 61.9 |
| 4 | InternVL3.5-8B | 61.9 |
| 5 | GPT-5 Chat | 59.5 |
| 6 | Gemma 3 4B (IT) | 52.88 |
| 7 | DeepSeek V3.2 | 47.03 |
Interactive version: theaggregate.ai/benchmark?slug=cvsbench-cvusa-satellite-to-ground · How It Works · Data refreshed daily, snapshot 2026-09-29.