Today we're releasing OlmoEarth v1.2, the latest in our family of open foundation models for Earth observation. 🌍
We've switched to rotary positional embeddings (RoPE), which reduces artifacts in the embeddings & gives a small performance boost. 🧵
OlmoEarth processes satellite images into tiles (patches), representing each as an embedding the model uses for downstream tasks—a numerical representation.
Earlier versions tagged patches with a fixed position signal that surfaced as unwanted artifacts in those embeddings.
V1.2 switches to RoPE. Instead of adding a position signal to each patch, it rotates the vectors the model compares in attention by angles defined by each patch's position.
Across all model sizes, we see consistent improvement on our kNN/linear-probe evals.
This update came directly from partners asking for cleaner embeddings. OlmoEarth v1.2 comes in Nano, Tiny, Small, & Base—all open source.
🤗 Models: https://t.co/9uGJLi9loK
💻 Training & fine-tuning code: https://t.co/QzncWyXuq6
📄 Tech report: https://t.co/bmeRA1zJwt
embedding — A list of numbers representing a piece of text's meaning, so that similar meanings end up numerically close and can be searched.
open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
Why it matters
OlmoEarth v1.2 swaps additive position signals for RoPE, clearing artifacts that showed up in the embeddingA list of numbers representing a piece of text's meaning, so that similar meanings end up numerically close and can be searched.Full definition → downstream users consume, with consistent probe gains across Nano through Base and open weightsA model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.Full definition → and training code.
