---
title: "OlmoEarth v1.1: A more efficient family of Earth observation models"
description: "OlmoEarth v1.1 is a new family of Earth observation models designed to reduce the compute required for large-scale satellite imagery predictions. Compute dominates the lifecycle from data export through preprocessing, inference, and post-processing, motivating a focus on transformer token sequence length. For Sentinel-2, the models collapse three resolution-specific tokens into one, reducing token counts threefold; this initially caused a 10 ppt drop on the m-eurosat kNN benchmark, so the pre-training regimen was modified. OlmoEarth v1.1 runs up to three times cheaper than v1 while maintaining similar performance across a mix of research benchmarks and partner-built tasks, although some regressions remain, and the same dataset is used to isolate methodological effects."
---

# OlmoEarth v1.1: A more efficient family of Earth observation models

[Hugging Face](https://yomu.fyi/company/hugging-face) · Kyle Wiggers · May 19, 2026

**Type:** Announcement

## Summary

OlmoEarth v1.1 is a new family of Earth observation models designed to reduce the compute required for large-scale satellite imagery predictions. Compute dominates the lifecycle from data export through preprocessing, inference, and post-processing, motivating a focus on transformer token sequence length. For Sentinel-2, the models collapse three resolution-specific tokens into one, reducing token counts threefold; this initially caused a 10 ppt drop on the m-eurosat kNN benchmark, so the pre-training regimen was modified. OlmoEarth v1.1 runs up to three times cheaper than v1 while maintaining similar performance across a mix of research benchmarks and partner-built tasks, although some regressions remain, and the same dataset is used to isolate methodological effects.

## Context

Running OlmoEarth over tens to hundreds of thousands of square kilometers makes compute the largest cost across data export, preprocessing, inference, and post-processing. Lower compute costs are intended to support more partners and make self-managed deployments faster and less expensive.

## Approach / What changed

OlmoEarth v1.1 reduces transformer sequence lengths by merging Sentinel-2's three resolution-specific tokens into one. Because naive merging reduced benchmark performance, the pre-training regimen was modified. The models use the same dataset as OlmoEarth v1 to isolate the effects of these methodological changes.

## Takeaways

- Sentinel-2 inputs previously produced one token per timestep at each of three resolutions; collapsing those resolutions produces three times fewer tokens.
- Naively merging resolution tokens caused a 10 ppt drop on the m-eurosat kNN benchmark, requiring changes to the pre-training regimen.
- OlmoEarth v1.1 is available in Base, Tiny, and Nano sizes and can run up to three times cheaper than OlmoEarth v1, with some reported regressions.

**Tags:** [Machine Learning](https://yomu.fyi/topic/machine-learning), [Open Source](https://yomu.fyi/topic/open-source), [Performance](https://yomu.fyi/topic/performance)

- Source: [Hugging Face](https://huggingface.co/blog/allenai/olmoearth-v1-1)
- Source URL: https://huggingface.co/blog/allenai/olmoearth-v1-1
- Ingested by Yomu: 2026-08-29T01:49:13.955Z

[Read original post](https://huggingface.co/blog/allenai/olmoearth-v1-1)