🔍 Read the full analysis: What Makes OlmoEarth V1.3's Planet-Wide Embeddings Stand Out? on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
The OlmoEarth team has released v1.3 of its satellite embedding model, annual global data at 10-meter resolution and a browser-based explorer. The team says its 128-dimensional embeddings ranked highest on 20 of 27 evaluations, but those results have not been independently verified in the source material.
The OlmoEarth team has released version 1.3 of its satellite embedding model, together with annual global embeddings at 10-meter resolution and a browser-based map explorer, as described in the original analysis. The team says the model ranked highest on 20 of 27 evaluations, but those comparisons are developer-reported and have not been independently verified in the source material.
An embedding is a compact numerical representation learned from imagery. OlmoEarth v1.3 processes yearly time series from Sentinel-1 radar and multispectral imagery from Sentinel-2 and Landsat, producing a 128-number vector for each 10-meter pixel. The first 64 dimensions can be used as a smaller representation. The precomputed global data is available through Source Cooperative, according to the release description.
The accompanying OlmoEarth Embeddings Explorer lets users view the data as false-color maps, switch between years and compare results with underlying Sentinel-2 imagery, with OlmoEarth Studio’s custom embeddings supporting AI innovation. Users can select a location to find visually similar places or add labeled points and train a small classifier, the team says. The tool is presented as usable without a sign-in or advanced technical knowledge.
In comparisons with AlphaEarth Foundations and Tessera v2, OlmoEarth’s 128-dimensional embeddings ranked highest in 20 of 27 evaluations; its 64-dimensional version led in 18. The team says the evaluations covered six continents and multiple tasks. Its technical report contains the methods and full results. The release also introduces PLANTEUR, a crop-type benchmark focused on tropical farming; the team says its embeddings generalized best to locations not seen by the classifier during training.
Reusable Features Could Cut Mapping Costs
The release offers researchers and other users a way to start some land-cover or change-mapping work without downloading and processing large satellite archives for every project. With precomputed pixel-level embeddings, a user can train a smaller classifier for a task or compare representations across years. That could reduce storage and computing demands, especially for teams without dedicated GPU resources.
Those potential savings do not make the resulting maps self-validating. The reported evaluation counts suggest the model may perform well across varied tasks, but practical value will depend on the location, labels and purpose of a specific project. Users still need to check outputs against imagery and other reliable evidence before drawing conclusions about land use or its causes.
The developers say the model weights, pretraining data and code are available. They estimate that reproducing the final pretraining run for OlmoEarth v1.3 Base would cost about $2,720 in rented H100 computing time. That is the team’s estimate, not a guaranteed price; actual costs can vary by provider, availability and setup.
satellite imagery analysis software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Earlier OlmoEarth use centered on fine-tuning through OlmoEarth Studio: users trained a model for a particular task and ran it over satellite imagery to produce a map. The team says that workflow requires GPU access and the capacity to download and store imagery. Embeddings offer another approach: process the imagery once, then reuse the resulting numerical representations across different mapping tasks.
The team says v1.3 used 53% of the computing time of v1.2 and 18% of the time used to train v1. It also reports emissions equal to 50% of v1.2 and 22% of v1, and says v1.3 matched v1.2 in fine-tuning performance across research and partner labels. These are the developers’ comparisons; the supplied material does not detail the emissions accounting.
A release example applies the embeddings to mining-related landscape change in Peru’s Madre de Dios region, including areas near La Pampa across 2017–2020 and 2020–2025. The source cites Amazon Conservation’s MAAP monitoring program and the 2019 Operation Mercurio crackdown for regional history. A change in embeddings can indicate landscape change, but it does not by itself establish that mining caused it.
““OlmoEarth’s 128-dimensional embeddings score highest on 20 of 27 evals and rank highest overall.””
— The OlmoEarth team
As an affiliate, we earn on qualifying purchases.
Independent Tests and Data Limits
The supplied material does not include an independent replication of the 27 evaluations. The team’s technical report provides methods and results, but the evaluation task selection and performance differences across regions will matter when judging how well the model fits a particular use. The reported aggregate rankings do not establish that it will lead on every task or in every location.
The release description does not specify how often the annual embeddings will be updated, what gaps or quality limits apply in all locations, or what access and computing requirements users may face when working at large scale. In the Peru example, identifying change in a map is distinct from confirming its cause. Attribution to mining, recovery or another process requires checking other evidence, such as satellite imagery, field observations or monitoring records.
As an affiliate, we earn on qualifying purchases.
Testing the Embeddings in Use
Researchers and other users can access the precomputed embeddings through Source Cooperative and try the browser explorer, according to the release description. The team’s technical report is the next reference for readers seeking the evaluation setup and full results. Practical testing across regions, tasks and labels will help show whether the reported advantages hold outside the developers’ comparisons.
The supplied material gives no schedule for future data updates or independent evaluations. Users applying the maps to decisions about land use or environmental change will need to validate findings with relevant imagery and monitoring evidence. The next useful milestone is evidence from those real-world applications, including how well results transfer to locations and conditions not represented in training or evaluation.
land cover classification software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did OlmoEarth release?
The team released OlmoEarth v1.3, annual global satellite embeddings at 10-meter resolution and a browser-based explorer for viewing and working with those data.
What does an OlmoEarth embedding represent?
It is a numerical representation learned from satellite imagery. V1.3 produces a 128-number vector for each 10-meter pixel; the first 64 dimensions can also be used as a smaller representation.
How did the model perform in the reported evaluations?
The OlmoEarth team says the 128-dimensional version ranked highest in 20 of 27 evaluations, while the 64-dimensional version led in 18. These are the developers’ results and have not been independently verified in the supplied material.
Can the embeddings prove what caused a landscape change?
No. Embeddings may help identify changes between periods, but a map alone does not establish the cause. Users need to check findings against imagery and other evidence, such as field data or monitoring records.
Where can users explore the data?
The team says the precomputed global embeddings are available through Source Cooperative. Its browser-based explorer can display false-color maps, compare years and show underlying Sentinel-2 imagery.
Primary source: Hugging Face · via ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
