Clay
Clay FoundationClay is an Earth-observation foundation model that embeds satellite imagery together with location, time, and sensor metadata. The release provides pretrained weights and a configurable training interface for representation learning and downstream geospatial tasks. The Clay Foundation develops and distributes it.
Verified 2026-09-10 via the Hugging Face model card and Clay Foundation repository.
Openness
4 high confidence- license
- Apache-2.0(OSI)
- weights
- open(Apache-2.0 checkpoint)
- data
- documented-not-released(source imagery and metadata schema documented
- code
- open(trainer, data module, configs, and model implementation released)
The full training machinery is public and the corpus is documented but not redistributed, matching the ladder's documented-data plus open-code rung.
- https://raw.githubusercontent.com/Clay-foundation/model/main/README.md recorded 2026-09-10
States that code and trained weights use Apache, and documents trainer.py, ClayDataModule, and configuration-driven training.
Adoption
1 high confidence107 downloads in the trailing 30 days for the declared checkpoint.
- https://huggingface.co/api/models/made-with-clay/Clay recorded 2026-09-10
downloads: 107; gated: false; private: false.
Capability
3 high confidenceTwo directly readable attributes put Clay one rung below Prithvi-EO; absent claims are not inferred from the architecture.
- https://raw.githubusercontent.com/Clay-foundation/model/main/README.md recorded 2026-09-10
Documents a configurable training path and sensor, time, location, and band metadata consumed by the model interface.
Verified 2026-09-10