Multi-Modal Foundation · allenai
OlmoEarth v1.2 Base
An Earth-observation foundation model for analysing Sentinel-1, Sentinel-2, and Landsat satellite imagery.
What OlmoEarth v1.2 Base is
OlmoEarth v1.2 Base is an Earth-observation foundation model developed by the Allen Institute for AI (Ai2). Designed for applications in earth science, climate, and conservation, the model generates feature representations from multi-sensor satellite imagery.
Serving as a base-sized checkpoint within the OlmoEarth family, it is designed for remote sensing workflows ranging from raw feature extraction to downstream task adaptation across single scenes and temporal image sequences.
How it works
The model is built on a Vision Transformer Base (ViT-Base) architecture with 114 million parameters. It accepts satellite inputs from Sentinel-1 radar, Sentinel-2 optical, and Landsat sensors, supporting both individual captures and image time series for image feature extraction.
Key facts
- Architecture
- Vision Transformer Base (ViT-Base)
- Parameter count
- 114M
- Supported sensors
- Sentinel-1, Sentinel-2, Landsat
- Primary task
- Image feature extraction
What you can use it for
- Satellite embedding computation
- Extracting feature vectors from optical and radar satellite scenes or image time series for downstream analysis.
- Downstream image segmentation
- Fine-tuning the model backbone to delineate spatial features and land classes in Earth-observation data.
- Multi-sensor Earth monitoring
- Analysing environmental and climate phenomena using aligned Sentinel-1, Sentinel-2, or Landsat imagery.
Run it here
We list this model but cannot run it yet: no library and no architectures — nothing can load it.
Everything on this page still applies. Running it needs a loader for this repository, not a bigger machine.
Strengths and limits
Good at
- Operates over multiple common Earth-observation platforms, including Sentinel-1, Sentinel-2, and Landsat.
- Supports both individual satellite scenes and multi-temporal image series.
- Provides pre-trained visual representations ready for downstream fine-tuning.
Watch out for
- Requires task-specific fine-tuning or output heads for specialised tasks such as semantic segmentation.
- Sensor compatibility is specified only for Sentinel-1, Sentinel-2, and Landsat imagery.
Inputs and outputs
You provide
Image
PNG, JPEG or TIFF. Sent to the model as-is.
You get back
Mean-pooled representation, one vector per input.
Licence
Check the licence before commercial use · other
The other licence carries terms worth reading before you depend on it.
Questions
Can I use OlmoEarth v1.2 Base commercially?
Check the licence before commercial use. The other licence carries terms worth reading before you depend on it.
What sensor data can this model process?
The model processes optical and radar imagery from Sentinel-1, Sentinel-2, and Landsat satellites.
Does the model support temporal data?
Yes, it supports both individual static satellite images and temporal image series.
How large is OlmoEarth v1.2 Base?
The model has 114 million parameters based on the ViT-Base architecture.
How to cite it
OlmoEarth v1.2: A more efficient family of OlmoEarth models. arXiv:2605.20804.
Written from OlmoEarth v1.2 Base's own documentation (model card). Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.