Agentic Geospatial Benchmarks

SpatialAgentBench

Reference benchmarks for agentic AI models in geospatial data processing. Developed based on nearly 30 years of experience in spatial analysis – transparent, reproducible, and practical.

These benchmarks evaluate agentic capabilities of AI models in solving complex spatial problems. They run on the Enterprise edition of our Spatial Agents – the full MCP server stack with QGIS, GRASS GIS, OpenStreetMap, and Planetary Computer. This makes them a practical reference standard for agentic work with GIS tools and geodata. The portfolio is continuously expanded with new models and task scenarios.

Benchmark Results

Model Scores Comparison Click to enlarge
Performance Rating by Category Click to enlarge

Scoring System

🥇 Gold ≥ 90 🥈 Silver ≥ 75 🥉 Bronze ≥ 60 ✔ Passed ≥ 40 ✘ Failed < 40

Detailed Results

Loading benchmark data…

Benchmark Overview