PanoSpace-Bench - Observer Distance: leaderboard
Observer distance on a 360-degree panorama (PanoSpace-Bench, from the PanoWorld paper by Wang et al., 2026): given one 360-degree equirectangular panorama, the model picks which listed object is physically closest to the observer in 3D. 250 three- or four-option multiple-choice questions scored by exact choice accuracy, in percent; higher is better.
Metric: Accuracy (%; 250 three- or four-option multiple-choice questions on one ERP panorama). Source: arxiv.org. Saturation forecast: Around 2029. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | MiMo-V2.5 | 51.6 | #353 |
| 2 | Ministral 3 14B | 49.6 | #1852 |
| 3 | Qwen 2.5 VL 7B Instruct | 48.8 | #1842 |
| 4 | Qwen 3.5 9B | 48.6 | #803 |
| 5 | Ministral 3 8B | 47.6 | #2036 |
| 6 | GPT-4o | 45.6 | #741 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=panospace-bench-observer-distance · How It Works · Data refreshed daily, snapshot 2026-11-12.