Hallucination detection benchmark for geospatial AI
How well AI models do hallucination detection on satellite, aerial and map data: 2 sets with results, scored the same way for every model.
Sets
| Set | Index | Tasks of this type | Best score on the set | Models |
|---|---|---|---|---|
| EVE hallucination | Spatial Intelligence Index | all | 68 | 10 |
| EVE hallucination detection | Spatial Intelligence Index | all | 86 | 11 |
Top models
Mean of each model's best score on the 2 sets, 0 to 100, for the 9 models that have run all of them.
| Rank | Model | Developer | Score |
|---|---|---|---|
| 1 | Claude Opus 5.5 | Anthropic | 77 |
| 2 | Claude Sonnet 5.5 | Anthropic | 76 |
| 3 | Nemotron 3.5 Super VL 120B | NVIDIA | 75 |
| 4 | Step 5 Preview | StepFun | 74 |
| 5 | Claude Haiku 5.5 | Anthropic | 74 |
| 6 | GLM-5.3-Flash | Z.ai | 74 |
| 7 | Nemotron 3 Ultra | NVIDIA | 73 |
| 8 | Nemotron 3.5 Lightning 30B | NVIDIA | 71 |
| 9 | Nemotron 3 Nano Omni 30B | NVIDIA | 63 |
Every task type · Every evaluation · Methodology
Frequently asked questions
Which AI model is best at hallucination detection on geospatial data?
Claude Opus 5.5 leads, on the sets below. Mean of each model's best score on the 2 sets, 0 to 100, for the 9 models that have run all of them.
What is this hallucination detection benchmark?
Every task in our sets that asks for hallucination detection, 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.
