Aurora

Forecasts global weather at 0.25° resolution in six-hour steps.

Field
Weather & climate
Method
Deep learning
Data it takes
gridded-fields
License
MIT

Aurora is Microsoft’s data-driven global weather model. Started from ERA5 reanalysis or IFS HRES analysis fields, it steps a global state forward six hours at a time at 0.25° resolution.

Upstream describes the release as research code rather than an operational forecast service. The higher-resolution, wave and air-pollution fine-tunes need initial conditions the lab does not hold, so the 0.25° forecast is the one available here.

The code and all released checkpoints are MIT-licensed.

Upstream project → · Paper →

Catalog entry last checked 2026-10-03. All models →

Run Aurora on your data

The lab runs this on its own compute, in a container, with the inputs and parameters recorded alongside the result. Initial scoping conversations are free.

Contact the Lab