AI BriefingAi2 Releases OlmoEarth v1.3 with State-of-the-Art Global Satellite Embeddings
·2026.10.23 01:11
Key point
The model generates 128-dimensional vectors for every 10-meter pixel on Earth, outperforming AlphaEarth Foundations and Tessera v2 across six continents.
Details
- Ai2 released OlmoEarth v1.3, a satellite embedding model that produces 128-dimensional vectors for every 10-meter pixel on Earth using Sentinel-1, Sentinel-2, and Landsat data.
- The model achieved state-of-the-art results, scoring highest on 20 of 27 evaluations against AlphaEarth Foundations and Tessera v2, while the 64-dimensional version led on 18 of 27.
- Precomputed annual embeddings for the entire planet are available on Source Cooperative, allowing users to train lightweight classifiers locally without GPU access or fine-tuning the base model.
- Training costs dropped significantly, with v1.3 requiring 53% of the compute time and 50% of the carbon emissions of v1.2, and 18% and 22% of v1 respectively.
- A new browser-based OlmoEarth Embeddings Explorer allows users to visualize embedding changes over time, find similar locations, and train classifiers on the spot without an account.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
CopilotKit Releases OpenIntelligentUI, an Open-Source Alternative to ChatGPT Intelligent UIArtCraft releases open-source IDE for AI image and video creation with 62-model catalog