[](https://huggingface.co/Ai2Comms)🌍 Learn more about the OlmoEarth Platform: https://allenai.org/olmoearth
OlmoEarth models are our suite of foundation models for Earth observation, pre-trained on approximately 10TB of multimodal satellite data. Governments, NGOs, and other mission-driven organizations are already applying OlmoEarth to use cases like deforestation monitoring, food security, and wildfire risk assessment.
At Ai2, we know how to train and release powerful open-source models. For organizations with robust engineering teams, an open-source model is enough to get up and running. However, most organizations in the environmental sector—the very ones best positioned to apply these models—lack the infrastructure or engineering teams to manage the entire lifecycle: data labeling, model fine-tuning, and large-scale inference. For over a decade, we have operated platforms like Skylight and EarthRanger, software that users worldwide rely on daily, meaning it must work every single day. This experience has taught us what it takes to make a real impact: running models cost-effectively at the right time and place, monitoring performance, turning raw outputs into actionable insights, and validating that these outputs drive the outcomes our partners want to achieve.
That is why we built the OlmoEarth Platform: an infrastructure suite designed to take geospatial models from fine-tuning and evaluation all the way to large-scale inference.
Inference at this scale introduces a unique set of challenges. Satellite imagery must be discovered and accessed across multiple data providers, aligned across different projections and resolutions, and processed efficiently. The results must then be stitched into geographically consistent maps, all while the infrastructure remains resilient to the common failures of distributed computing.
Today,