Krux
May 21, 2026
OlmoEarth v1.1 Cuts Satellite AI Costs by 3x
Published: May 21, 2026 at 12:23 AM
Updated: May 21, 2026 at 12:23 AM
100-word summary
AI models that analyze satellite imagery just got three times cheaper to run. OlmoEarth v1.1, released May 19, slashes compute costs by rethinking how it processes Earth observation data. The trick? Smarter tokenization that splits satellite images into fewer pieces without losing accuracy. What can you do with that? Organizations already tracking mangrove loss or mapping crops across entire countries can now refresh those maps three times more often on the same budget. Or train custom models for new environmental monitoring projects in a fraction of the time. The catch: some tasks take a performance hit, though the team says most uses hold steady.
What happened
AI models that analyze satellite imagery just got three times cheaper to run. OlmoEarth v1.1, released May 19, slashes compute costs by rethinking how it processes Earth observation data. The trick? Smarter tokenization that splits satellite images into fewer pieces without losing accuracy. What can you do with that? Organizations already tracking mangrove loss or mapping crops across entire countries can now refresh those maps three times more often on the same budget. Or train custom models for new environmental monitoring projects in a fraction of the time.
Why it matters
The catch: some tasks take a performance hit, though the team says most uses hold steady.