Visual question answering benchmark for geospatial AI
How well AI models do visual question answering on satellite, aerial and map data: 5 sets with results, scored the same way for every model.
Sets
| Set | Index | Tasks of this type | Best score on the set | Models |
|---|---|---|---|---|
| Vision (no tools) | Spatial Intelligence Index | all | 59 | 24 |
| BigEarthNet.txt | Spatial Intelligence Index | all | 60 | 16 |
| GeoChrono | Spatial Intelligence Index | all | 76 | 16 |
| GeoChrono larger set | Spatial Intelligence Index | all | 69 | 5 |
| DVL-Bench | Spatial Intelligence Index | all | 31 | 8 |
Top models
Best score on Vision (no tools), the set most models have run, 0 to 100.
| Rank | Model | Developer | Score |
|---|---|---|---|
| 1 | Claude Opus 5.5 | Anthropic | 59 |
| 2 | Claude Sonnet 5.5 | Anthropic | 59 |
| 3 | GPT-5.6 Sol | OpenAI | 58 |
| 4 | GPT-6 Sol | OpenAI | 56 |
| 5 | GPT-6.1 Sol | OpenAI | 56 |
| 6 | GPT-5.6 Terra | OpenAI | 54 |
| 7 | Claude Opus 5 | Anthropic | 53 |
| 8 | Qwen 3.8 27B | Alibaba (Qwen) | 53 |
| 9 | GPT-6 Luna | OpenAI | 53 |
| 10 | Claude Haiku 5.5 | Anthropic | 52 |
Every task type · Every evaluation · Methodology
Frequently asked questions
Which AI model is best at visual question answering on geospatial data?
Claude Opus 5.5 leads, on the sets below. Best score on Vision (no tools), the set most models have run, 0 to 100.
What is this visual question answering benchmark?
Every task in our sets that asks for visual question answering, in GIS and Earth observation work, with each set's own score. It is a view over the sets, not an index of its own.
Which index does it count in?
Each set counts in one index: the Spatial Intelligence Index, as the table shows.
