Ai2 has introduced the OlmoEarth Platform, infrastructure for using its Earth observation foundation models in practical geospatial projects. The platform is meant to help organizations move from open satellite models to fine-tuning, evaluation, and large-scale inference.
OlmoEarth models were pretrained on roughly 10 terabytes of multimodal satellite data. Ai2 says governments, NGOs, and mission-driven organizations are already adapting them for work such as deforestation monitoring, food security, and wildfire risk.
The platform addresses a common gap in applied AI. Many environmental organizations may benefit from geospatial models but do not have teams to manage labeling data, tuning models, running inference at planetary scale, and monitoring results over time.
Ai2’s argument is that open weights alone are not enough for impact in this field. The hard part is reliable operations: running models at the right time and place, turning outputs into usable signals, and checking whether those signals improve decisions on the ground.