UrbanWell - Land Use Classification: leaderboard
UrbanWell (2026) tests multimodal models on urban analytics from satellite and street-view images of grid cells in 38 cities, most of them European. This board is the accuracy in percent of choosing the land-use category of a cell from same-year images; higher is better.
Metric: Accuracy (%) of choosing the land-use category of a grid cell from same-year satellite and street-view images (multiple choice), cells sampled evenly across 38 mostly European cities, zero-shot; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.0 Flash | 66 | #704 |
| 2 | Gemma 3 4B | 60 | #5064 |
| 3 | Gemma 3 27B | 59 | #1631 |
| 4 | GPT-5 Nano | 58 | #889 |
| 5 | Gemma 3 12B | 55 | #2000 |
| 6 | Mistral Small 3.1 | 52 | #1689 |
| 7 | Qwen 2.5 VL 7B Instruct | 51 | #1833 |
| 8 | Nova Lite (v1) | 49 | #2451 |
| 9 | Qwen 2.5 VL 32B Instruct | 48 | #1053 |
| 10 | Phi-4 Multimodal Instruct | 25 | #3882 |
| 11 | Llama 4 Scout | 7 | #1776 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=urbanwell-land-use-classification · How It Works · Data refreshed daily, snapshot 2026-11-13.