According to GeekWire, the Allen Institute for AI announced OlmoEarth on Tuesday as an open, end-to-end system that uses AI models trained on millions of Earth observations totaling roughly 10 terabytes of data. The platform includes OlmoEarth Studio for creating datasets and fine-tuning models, plus OlmoEarth Viewer as a web app for exploring AI-generated maps. Early adopters are already using it to update global mangrove maps twice as fast with 97% accuracy, detect deforestation across the Amazon, and map vegetation dryness in Oregon for wildfire prediction. Ai2 CEO Ali Farhadi said the initiative aims to make Earth AI accessible to those working on the front lines of environmental challenges, while lead researcher Patrick Beukema emphasized collaboration across scientific fields.
Sponsored content — provided for informational and promotional purposes.
Democratizing what was once exclusive
Here’s the thing about geospatial analysis: it’s been dominated by Big Tech for years. Google Earth Engine and Microsoft’s Planetary Computer have provided access to petabytes of satellite data, but they’re not exactly beginner-friendly. You need serious technical chops to actually do anything useful with them. And they’re not fully open-source, which means you’re working within their walled gardens.
Now Ai2 is coming in with a completely different approach. They’re giving away the whole stack – the models, the tools, the infrastructure. Basically, they’re saying “here’s everything you need to understand what’s happening to the planet, no strings attached.” It’s the same philosophy behind their open-weight language models, just applied to climate science and conservation.
Beating the giants at their own game
What’s really interesting is how Ai2 is positioning OlmoEarth against the competition. They’re directly comparing their models to Google’s AlphaEarth Foundations and claiming their fine-tuned versions “outperformed AEF substantially.” They’re also stacking up well against models from Meta, IBM, and NASA.
But here’s the key difference: Google released “embeddings” rather than the actual model. That’s like giving you the answers without showing the work. Ai2 is giving researchers the actual tools to build and customize their own solutions. It’s a fundamentally different approach to AI development – one that prioritizes collaboration over control.
Real impact, right now
The early results are pretty compelling. Updating mangrove maps twice as fast with 97% accuracy? That’s not some theoretical improvement – that’s meaningful progress for conservation efforts. Detecting deforestation across the Amazon and mapping vegetation dryness for wildfire prevention? These are exactly the kinds of applications that could make a difference in real environmental crises.
And the timing couldn’t be better. With climate change accelerating and environmental monitoring becoming increasingly critical, we need more eyes on the problem. The OlmoEarth Viewer is available starting today, and they’ve already put the code and documentation on GitHub. The full platform with OlmoEarth Studio is rolling out to select partners, and they’re actively seeking more collaborations.
Bigger than just another tool
This feels like part of a larger shift in how we approach AI for social good. Instead of keeping the most powerful tools locked up in corporate labs, organizations like Ai2 are pushing for what they call “true openness.” It’s not just about access to data – it’s about giving people the complete system to build upon.
Think about who benefits here: smaller conservation groups, academic researchers, local governments. Organizations that don’t have Google-level resources but are doing critical work on the ground. By leveling the playing field, Ai2 might actually accelerate our ability to understand and respond to environmental challenges. And honestly, that’s exactly what open science should be about.
