Field noteOct 22, 20261 min read1 source

OlmoEarth v1.3: State-of-the-art embeddings for the whole planet

In one minute

AI2 has released OlmoEarth v1.3, an open satellite-embedding model, alongside precomputed annual embeddings and the browser-based OlmoEarth Embeddings Explorer.

What changed: Same event as OlmoEarth v1.3: State-of-the-art embeddings for the whole planet

The release is available under the OlmoEarth Artifact License with model weights, pretraining data, and code.

The model produces a 128-dimensional vector for every 10-meter pixel from yearly Sentinel-1 radar plus Sentinel-2 and Landsat multispectral time series; its first 64 dimensions can also be used separately. AI2 says the 128-dimensional version ranked highest on 20 of 27 evaluations against AlphaEarth Foundations and Tessera v2, while the 64-dimensional version led 18. It estimates the final pretraining run at roughly $2,720 of rented H100 time.

The Explorer requires no account and can compare years, find visually similar locations, train a small classifier from labeled points, and export GeoTIFFs. Global coverage is available for 2023, 2024, and 2025; 2017 through 2022 cover selected countries. AI2 says the archive is about 150 TB per year and includes source-scene provenance per tile. The source does not establish independent benchmark replication or accuracy for a particular mapping application or region.

Evidence / sources

  1. 1huggingface.co/blog/allenai/olmoearth-v1-3
All Field Notes

Next reads

Read next

Open the graph
  1. 01topicOpen ModelsExplore the open-models topic hub.
  2. 02topicModelsExplore the models topic hub.
  3. 03topicRetrievalExplore the retrieval topic hub.
  4. 04archiveField NotesOpen the latest editorial analysis.
  5. 05source ledgerSource LedgerInspect the public source evidence graph.
Who read this page?Machine requests, hidden until opened

Loading the privacy-safe route aggregate...

Open the JSON contract