Ai2 ships open satellite embedding model OlmoEarth v1.3 with free global map explorer
In short: Ai2 has published OlmoEarth v1.3, an embedding model that condenses a full year of radar and multispectral satellite data into a single compact vector for every 10-meter patch of the planet. Alongside it comes a free browser tool, the OlmoEarth Embeddings Explorer, that lets anyone compare landscapes across years or train tiny classifiers without touching a GPU. Benchmarked against AlphaEarth Foundations and Tessera v2 across 27 tests, the new model comes out on top overall while costing far less to train than its predecessors.
This summary was generated automatically by AI from Hugging Face's publication. It is our own text, not a copy of the original — facts, figures and quotes belong to the source, linked above and below.
What changed?
- 1Encodes a year of Sentinel-1, Sentinel-2 and Landsat observations into one 128-dimensional embedding per 10m pixel; a lighter 64-dimensional version also works on its own
- 2Global precomputed embeddings released publicly through Source Cooperative, with an archive of roughly 150 TB per year
- 3Launch of the free OlmoEarth Embeddings Explorer, enabling year-over-year comparisons, similarity search and on-the-spot classifier training with no account or GPU needed
- 4Leads AlphaEarth Foundations and Tessera v2 on 20 of 27 benchmark tasks at 128 dimensions, and 18 of 27 at 64 dimensions
- 5Adds PLANTEUR, a new benchmark for tropical crop-type classification
- 6Training required only 53% of v1.2's compute and 18% of the original v1's, cutting emissions to roughly half and a fifth of those versions respectively
- 7Reproducing the final training run is estimated to cost about $2,720 in rented H100 time
- 8Model weights, pretraining data and code remain fully open under the OlmoEarth Artifact License
| Parameter | Before | Now |
|---|---|---|
| Output | Not specified | One 128-dimensional vector per 10m pixel, derived from a year of Sentinel-1, Sentinel-2 and Landsat time series; a 64-di |
| Benchmark performance | OlmoEarth v1.2 | Tops AlphaEarth Foundations and Tessera v2 on 20 of 27 evaluations (128-dim) and 18 of 27 (64-dim) |
| Training compute | v1 baseline | v1.3 used 53% of v1.2's training time and 18% of v1's, cutting carbon output to 50% and 22% of those versions respective |
| Reproduction cost | Not specified | Roughly $2,720 in rented H100 time for the final training run |
| Licensing | OlmoEarth Artifact License (v1, v1.2) | Same open license applied to v1.3, weights, data and code |
Why it matters
It lowers the barrier for geospatial analysis — organizations can now build crop, land-cover or deforestation maps by training a lightweight classifier on ready-made embeddings instead of running costly fine-tuning jobs, and the explorer tool extends that capability to non-technical users entirely in the browser.
Sources
- Hugging FaceOfficialPrimary sourceOriginal article →„OlmoEarth v1.3: State-of-the-art embeddings for the whole planet“22 Oct 2026, 19:11
- Published by source
- 22 Oct 2026, 19:11
- Found by our system
- 22 Oct 2026, 19:29
- Summary generated
- 22 Oct 2026, 19:31
This article was written by AI from the original source. Facts, numbers and prices come from the source; missing values are marked “Not specified”. Legal notice, copyright and privacy