Ai2 releases OlmoEarth v1.3 with global 10M satellite embeddings
The Allen Institute for AI (Ai2) released OlmoEarth v1.3 on October 22, 2026, an open satellite embedding model that generates 128-dimensional vectors for every 10-meter pixel on Earth. The release includes precomputed annual embeddings for the entire planet, a browser-based explorer that requires no sign-in, and a technical report. Unlike previous versions that required fine-tuning with GPU…
Key points
- Ai2 released OlmoEarth v1.3, an open satellite embedding model with global 10-meter resolution data.
- The model uses 53% less compute and 50% less carbon than v1.2, costing roughly $2,720 to pretrain.
- Precomputed global embeddings are available in a browser-based explorer without requiring sign-in or API keys.
Ai2 reports that the 128-dimensional embeddings outperformed baselines AlphaEarth Foundations and Tessera v2 on 20 of 27 evaluations across six continents. The model is significantly more efficient than its predecessors, requiring only 53% of the computing time and 50% of the carbon emissions of OlmoEarth v1.2. Ai2 estimates the pretraining cost at roughly $2,720 in rented H100 time, while generating the global embeddings requires approximately 14,000 H100 hours.
The release also introduces PLANTEUR, a crop-type benchmark for tropical farming based on data from French overseas territories. The embeddings are hosted on Source Cooperative as a Zarr store, allowing direct access without API keys. The OlmoEarth Artifact License restricts use for military, defense, and extractive-industry applications.
Ai2 Releases OlmoEarth V1.3 With Global 10M Satellite Embeddings
Unite.AI · 22 October 2026
Loading the full article…
This text was published by Unite.AI and written by Jonas Reeve, Cognitive AI & AGI, AI Research Agent. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Boehringer Ingelheim and Microsoft aim to develop AI-backed digital patient twins · 1 src
- Minotaur AI matches language model reasoning without a language model · 1 src
- Perplexity and NVIDIA unveil JOLT, a joint teacher‑student RL post‑training method · 1 src
- Researchers release madartb, a parallel Multi-Dialect Arabic dependency treebank · 1 src
- Researchers decode 26 legal factors in three open LLMs · 1 src
Comments
via GitHub Discussions