GLEN: Geospatial LLM Enabled Navigator

GLEN is open source software from the Eric and Wendy Schmidt Center for Data Science and Environment at UC Berkeley. It lets people ask questions of public conservation data in plain language and get back maps, tables, and charts, along with the queries and data sources behind them.

We asked the live California 30x30 app, in October 2026, to show bird species richness as 3D hexagons, using the open weight model Kimi K3. The agent reads the ACE biodiversity dataset's schema, turns its native bird counts into a hexagon layer, draws it with height and color both showing the number of species, and adds a legend. We then turn and tilt the map with the keyboard.

About

Accelerating conservation decisions with open source AI.

Conservation decisionmakers often need answers that draw on dozens of geospatial datasets, quickly. GLEN connects a language model to authoritative data and open source geospatial tools, so anyone can get fast, reproducible analysis.

  • Talk to the dataAsk questions in plain language and get answers on the map.

    Type a question, or draw an area on the map, and GLEN finds the data, runs the analysis, and shows the answer as a map, a table, or a chart.

  • Answers others can checkEvery query can be shared, checked, and changed by others.

    GLEN shows each step it takes and points out the caveats of each dataset and the limits of what it can show, so that results are read for what they are.

  • Runs on a small server, not a data centerThat keeps cost and energy use low.

    Running on local hardware rather than in a commercial data center also lowers privacy concerns and keeps users in control of their data.

  • Fully open sourceAnyone can inspect, modify, and redistribute it.

    GLEN and the tools it is built from are open source under the Apache 2.0 license, and its data comes from public sources, each cited with its license.

Benefits of local language models

Open local language models offer many benefits for using AI in science. These models run locally on a user's computer (no data center required) and provide processing power comparable to commercial models.

Robust Science

Open local models offer researchers sufficient speed and quality of results.

Energy Efficient

Can use 100x less energy than a commercial model.

Privacy

Users have total control over their data.

Freedom of Use

While requiring some technical expertise, open local models can be viewed, used, modified, and distributed by anyone.

Low Cost

Open local models run for free.

Enhanced Reliability

Access to the internet or the cloud are not required.

Applications

GLEN is an open source library and can be used to build a decision-support tool for any species, region, or policy. The applications below are a few examples.

Partners

We build each application with organizations that bring the data, expertise, and questions behind it.

Team

  • Carl Boettiger

  • Cassie Buhler

  • Kristin Davis

  • Magali de Bruyn

  • Matt Fisher

  • Ciera Martinez

Contact

For press, partnerships, or support for the work, email us: dse@berkeley.edu