SpaceThinker: benchmark results
Provider: Other. Access: Open.
Aggregate Rating 1564 ± 45, rank #1496 of 10916 rated models, from 28 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SpatialThinker Release - HallusionBench | 65.4 | Accuracy (%) | 75 |
| SpatialThinker Release - 3DSRBench | 51.1 | Accuracy (%) | 66.7 |
| SpatialThinker Release - BLINKMultiView | 50.4 | Accuracy (%) | 58.3 |
| DriveSpatial - Understanding | 31.74 | Accuracy (%) on the multiple-choice spatial-understanding ta | 54.5 |
| SpatialThinker Release - BLINKSpatialRelation | 73.4 | Accuracy (%) | 47.2 |
| SpatialThinker Release - CV-Bench2D | 65.1 | Accuracy (%) | 46.7 |
| UAV-DualCog - Flight Behavior Recognition (Composite) | 16.1 | Composite-level behavior accuracy (%; 1,024 first-person UAV | 42.3 |
| SpatialThinker Release - CV-BenchAverage | 65.5 | Accuracy (%) | 40 |
| DriveSpatial - Construction | 27.16 | Accuracy (%) on the scene-construction tasks (building a coh | 36.4 |
| SpatialThinker Release - SpatialReasonerEval | 69.6 | Accuracy (%) | 33.3 |
| UAV-DualCog - Future Observation Prediction | 25.4 | Answer accuracy (%; 1,024 questions asking which view the UA | 29.4 |
| UAV-DualCog - Landmark Visibility Counting | 33.7 | Visibility-count accuracy (%; 1,024 flight videos in which t | 28.6 |
Interactive version: theaggregate.ai/model?slug=spacethinker · How It Works · Data refreshed daily, snapshot 2026-11-08.