Multi-Modal Foundation · allenai
OlmoEarth v1.1 Nano
A 5.5-million-parameter vision transformer for remote sensing analysis on satellite imagery and time series.
What OlmoEarth v1.1 Nano is
OlmoEarth v1.1 Nano is a compact Earth-observation foundation model developed by the Allen Institute for AI. It provides a minimal-footprint checkpoint intended for embedded devices and high-throughput data processing workflows across earth science, climate studies, and conservation research.
The model handles remote sensing data acquired by Sentinel-1, Sentinel-2, and Landsat sensors. It can ingest individual satellite images or multi-temporal image series, serving either as an encoder for computing dense spatial embeddings or as a backbone for fine-tuning on downstream tasks.
How it works
The model employs a ViT-Nano vision transformer architecture containing 5.5 million parameters. It processes optical and radar inputs from Sentinel-1, Sentinel-2, and Landsat, extracting representations from single captures or sequences of images across time.
Key facts
- Architecture
- ViT-Nano
- Parameters
- 5.5M
- Supported sensors
- Sentinel-1, Sentinel-2, Landsat
- Input modalities
- Images and image time series
What you can use it for
- Satellite image segmentation
- Fine-tune the model to delineate land cover, environmental boundaries, or surface changes from satellite scenes.
- Remote sensing embedding extraction
- Generate dense feature vectors from satellite images for spatial indexing, clustering, and downstream classification.
- Time series landscape monitoring
- Process temporal sequences of Sentinel-1, Sentinel-2, or Landsat imagery to observe environmental transitions over time.
- High-throughput and embedded Earth observation
- Execute remote sensing inference on resource-constrained edge hardware or high-volume observation pipelines.
Run it here
We list this model but cannot run it yet: no pipeline_tag — needs a human.
Everything on this page still applies. Running it needs a loader for this repository, not a bigger machine.
Strengths and limits
Good at
- A small footprint of 5.5 million parameters enables deployment in embedded environments and large-scale, high-throughput pipelines.
- Directly supports multiple satellite sensor types, including Sentinel-1, Sentinel-2, and Landsat.
- Accepts both single static scenes and multi-temporal image series as input.
Watch out for
- The 5.5M parameter capacity may offer lower representation capacity than larger transformer variants on complex scenes.
- Documented sensor compatibility is limited to Sentinel-1, Sentinel-2, and Landsat imagery.
- Application to specialised environmental goals requires task-specific downstream fine-tuning.
Inputs and outputs
You provide
Sequence or text
A protein or nucleotide sequence, or free text — whichever this model was trained on.
You get back
Mean-pooled representation, one vector per input.
Licence
Check the licence before commercial use · See model card
The See model card licence carries terms worth reading before you depend on it.
Questions
Can I use OlmoEarth v1.1 Nano commercially?
Check the licence before commercial use. The See model card licence carries terms worth reading before you depend on it.
What satellite instruments does OlmoEarth v1.1 Nano support?
The model is built to process observations from Sentinel-1 radar, Sentinel-2 optical sensors, and Landsat satellites.
Can the model evaluate satellite observations across time?
Yes, the architecture is designed to handle both individual satellite scenes and multi-image temporal series.
What is the parameter size and architecture of this model?
The model is based on the ViT-Nano architecture and consists of 5.5 million parameters.
How can the model be applied to downstream tasks?
It can be used directly to extract numerical feature embeddings or fine-tuned for downstream targets such as semantic segmentation.
Written from OlmoEarth v1.1 Nano'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.