What is LGM? A guide to the Large Geospatial Model

A location question is rarely just a question about coordinates. Choosing a site for a public market means understanding who lives nearby, how people reach it, which businesses already operate there and what the land can legally support. Each part of the answer lives in a different dataset.

LGM, short for Large Geospatial Model, brings those inputs into a shared geospatial intelligence platform. Its purpose is to help applications and AI agents work with territory: retrieve relevant data, perform spatial operations and return results that people can inspect on a map.

What does a Large Geospatial Model do?

In the LGM product, the name describes a platform that combines data services, spatial algorithms, AI orchestration and user interfaces. Its capabilities are delivered through a backend that can serve more than one application.

A GIS user might request a layer. An AI agent might ask for buildings inside a polygon. An analyst might investigate which neighborhoods need better access to services. These interactions can draw on the same data and computation services while returning different representations of the result.

This shared foundation matters when an organization already has a map, assistant or operational system. It can integrate the capabilities it needs without requiring every user to adopt the same interface.

How a territorial question becomes an investigation

LGM’s research workflow follows five steps:

1. Understand: identify the territory, decision and constraints.
2. Retrieve: find relevant datasets that the user is allowed to access.
3. Compute: perform the necessary spatial operations.
4. Verify: examine sources, coverage, assumptions and consistency.
5. Return: provide an explanation, indicators and map-ready geometry.

Consider a request to find underserved areas for neighborhood sports facilities. The investigation needs more than a list of existing facilities. It must consider population, practical access, administrative boundaries and data completeness. A useful output identifies candidate areas and explains what additional evidence could change their ranking.

Where the data comes from

LGM’s data foundation spans global sources, regional datasets and private collections. Public layers provide a starting point; local knowledge makes the investigation more specific.

For example, a building footprint can indicate where development exists. Municipal records may explain occupancy or planned changes. An organization’s own asset data can add operating constraints that public maps do not capture.

The quality of the conclusion depends on the coverage, age and suitability of these inputs. Missing records should be visible in the analysis, rather than silently interpreted as an absence of activity.

How people and applications use LGM

The Web GIS and chat provide a visual environment for investigation. The Platform manages account data, collections and API keys. Developers can explore API integration or connect an agent through the MCP Server. The QGIS integration offers a path into professional GIS work.

Specialized fine-tuned models and distribution through partner model catalogs are part of LGM’s development direction. Computer Vision is also being developed to turn observations from imagery into georeferenced objects and changes.

Start with a decision that can be checked

A productive first task has a defined territory, a concrete decision and a clear output. “Compare candidate areas for a small public market and list the missing evidence” is more useful than “analyze this city.”

The Tbilisi research illustrates how that approach narrows a broad urban question into specific spatial priorities. To explore your own territory, open LGM GIS.