SpatialUAV - Distance Comparison: leaderboard
Depth comparison in real low-altitude UAV images from SpatialUAV (Zhang et al., 2026): the model picks which marked region is closer to or farther from the drone, against answers derived from Metric3D depth estimates and checked by annotators. Exact-match accuracy in percent over 261 questions; higher is better.
Metric: Distance Comparison Accuracy (%; exact region ID of the closer or farther marked region). Source: arxiv.org. Saturation forecast: Around 2029. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.6 27B | 69.7 | #339 |
| 2 | Qwen 3.5 9B | 66.3 | #803 |
| 3 | Qwen 3.6 35B A3B | 65.5 | #461 |
| 4 | Claude Opus 4.7 | 60.2 | #75 |
| 5 | GPT-5.4 | 53.6 | #121 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=spatialuav-distance-comparison · How It Works · Data refreshed daily, snapshot 2026-11-12.