Multi-Modal Foundation · microsoft
Aurora
An Earth system foundation model for global weather forecasting, atmospheric chemistry, and cyclone tracking.
What Aurora is
Aurora is a foundation model for the Earth system developed by Microsoft. It is designed to model multi-domain planetary processes, spanning physics, climate, and earth science.
Rather than deploying specialised, independent systems for individual atmospheric phenomena, the model unifies global weather forecasting, atmospheric chemistry, ocean wave behaviour, and tropical-cyclone tracking within a single shared deep-learning backbone.
How it works
The model employs a deep-learning foundation architecture configured as a single shared backbone capable of processing multiple Earth system disciplines, including atmospheric dynamics and chemistry. Model weights are provided across various versions designed for Earth system evaluation.
Key facts
- Category
- Weather Foundation Model
- Architecture
- Single shared backbone foundation model
- Primary fields
- Climate, earth-science, physics
What you can use it for
- Global Weather Forecasting
- Simulating and predicting large-scale planetary atmospheric dynamics and operational weather conditions.
- Atmospheric Chemistry Prediction
- Modelling chemical processes, atmospheric constituents, and composition changes over time.
- Ocean Wave Forecasting
- Predicting surface ocean wave conditions across marine environments.
- Tropical-Cyclone Tracking
- Forecasting and tracking the paths and progression of tropical cyclones.
Run it here
We list this model but cannot run it yet: no pipeline_tag — needs a human.
Everything on this page still applies. Running it needs a loader for this repository, not a bigger machine.
Strengths and limits
Good at
- Unifies global weather forecasting, ocean waves, atmospheric chemistry, and cyclone tracking into one shared backbone.
- Provides a single foundation model architecture adaptable across diverse Earth science and physics domains.
Watch out for
- The provided documentation does not specify computational resource requirements or hardware dependencies for inference and fine-tuning.
- Spatial resolutions and temporal forecasting ranges are not detailed in the available repository materials.
Inputs and outputs
You provide
Sequence or text
A protein or nucleotide sequence, or free text — whichever this model was trained on.
You get back
Mean-pooled representation, one vector per input.
Licence
Check the licence before commercial use · mit
The mit licence carries terms worth reading before you depend on it.
Questions
Can I use Aurora commercially?
Check the licence before commercial use. The mit licence carries terms worth reading before you depend on it.
What tasks can Aurora perform?
Aurora can forecast global weather, model atmospheric chemistry, predict ocean wave dynamics, and track tropical cyclones.
Does Aurora require separate architectures for different climate components?
No, it addresses atmospheric dynamics, atmospheric chemistry, ocean waves, and cyclones through a single shared foundation backbone.
Are there different variants of the model available?
Yes, weights are provided for various versions of Aurora.
How to cite it
Cristian Bodnar, Wessel P. Bruinsma, Ana Lucic, Megan Stanley, Anna Allen, Johannes Brandstetter, Patrick Garvan, Maik Riechert, Jonathan A. Weyn, Haiyu Dong, Jayesh K. Gupta, Kit Thambiratnam, Alexander T. Archibald, Chun-Chieh Wu, Elizabeth Heider, Max Welling, Richard E. Turner, Paris Perdikaris. A Foundation Model for the Earth System. Nature, 2025. doi: 10.1038/s41586-025-09005-y
Written from Aurora's own documentation (model card). Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.