Ai2 Releases OlmoEarth v1.3 with Browser Map-Making Tool
The Allen Institute for Artificial Intelligence has launched OlmoEarth v1.3, providing global satellite embeddings and a new browser-based exploration tool.
The Allen Institute for Artificial Intelligence, known as Ai2, has released OlmoEarth v1.3, a state-of-the-art satellite map embedding model covering the entire planet. Alongside the model, the organization launched precomputed annual embeddings at various resolutions and a browser-based map-making tool called the OlmoEarth Embeddings Explorer.
The system condenses a full year of satellite data into a numerical fingerprint for every 10-by-10-meter patch of ground, with 64-square-meter and 128-meter resolutions also accessible. These fingerprints capture underlying patterns that allow software to differentiate between land cover types such as forests, mines, and crops. The model processes radar and optical imagery from missions including Sentinel-1, Sentinel-2, and Landsat, combining radar reflections with visible and infrared wavelength captures.
According to Ai2, version 1.3 is its most economical model to train to date. It required only 53% of the computing time needed for version 1.2 and 18% of the computing time of the initial version, achieving corresponding reductions in carbon emissions.
Practical Applications of Satellite Embeddings
Satellite embedding models help condense large amounts of spatial and temporal data to monitor environmental changes. For example, researchers can track land use shifts over time without needing to train custom models. Historical data from 2016 to 2020 clearly displays mining shifts and subsequent enforcement actions in Peru's Madre de Dios region following government intervention. Similarly, agricultural analysts can track the expansion of oil palm plantations and identify areas where forests may have been cleared.
Previously, utilizing OlmoEarth required fine-tuning within OlmoEarth Studio, which involved substantial graphics processing unit resources, downloading large imagery sets, and running task-specific training. The newly introduced browser-based explorer allows users to analyze regions and track changes in seconds without requiring sign-ins, local storage, or GPU access.
What it means for developers
For engineering teams, the release lowers the barrier to entry for geospatial analysis. Testing an idea against satellite data no longer demands an expensive, time-consuming fine-tuning run. Developers can perform a first-pass analysis in seconds using the explorer, helping them decide if a project warrants the compute time of full fine-tuning. For those working across diverse tech stacks, developers can try top AI models cheaply through one API at https://apixoai.online to build and scale data processing workflows efficiently.
Source: Ai2 releases OlmoEarth v1.3 state-of-the-art AI-ready satellite embeddings for the whole planet — SiliconANGLE AI. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key
