The OlmoEarth v1.1 family introduces a more efficient architecture for analyzing satellite imagery. These models leverage a masked autoencoder approach to process geospatial data with lower computational overhead. Researchers improved training stability and inference speed.
The Signal
This update allows practitioners to deploy high-resolution environmental monitoring tools on more modest hardware without sacrificing accuracy.