SpatialBot-3B: benchmark results

Provider: BAAI. Access: Open.

Aggregate Rating 1554 ± 45, rank #1656 of 10916 rated models, from 18 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
CVSBench - CVUSA Ground-to-Satellite68.3Accuracy (%) on CVUSA-subset ground-to-satellite questions: 71.4
SpatialThinker Release - CV-BenchDepth77.3Accuracy (%)66.7
SpatialThinker Release - BLINKRelativeDepth67.7Accuracy (%)63.9
CVSBench - CVUSA Satellite-to-Ground59.08Accuracy (%) on CVUSA-subset satellite-to-ground questions: 53.8
SpatialThinker Release - CV-Bench3D69.1Accuracy (%)52.6
CVSBench - FOV Satellite-to-Ground42.38Accuracy (%) on FOV-subset satellite-to-ground questions (li50
CVSBench - FOV Ground-to-Satellite29.18Accuracy (%) on FOV-subset ground-to-satellite questions (li38.5
SpatialThinker Release - CV-BenchDistance60.8Accuracy (%)38.1
SpatialThinker Release - BLINKAverage67.8Accuracy (%)36.1
SpatialThinker Release - CV-BenchRelation69.4Accuracy (%)31.6
CVSBench - Viewpoint Arrow26.5Accuracy (%) on View-Arrow: choose the camera-pose arrow on 28.6
SpatialThinker Release - BLINKSpatialRelation67.8Accuracy (%)27.8

Interactive version: theaggregate.ai/model?slug=spatialbot-3b · How It Works · Data refreshed daily, snapshot 2026-11-08.