Microsoft releases Aurora 1.5 as an open weather and Earth-system AI model

Microsoft Research has released Aurora 1.5, a major update to its Aurora Earth-system foundation model, aiming to make AI-based weather intelligence more useful for researchers, developers, and organizations that depend on short-term atmospheric forecasts.

The July 9 announcement says Aurora 1.5 adds 22 weather variables, hourly temporal resolution, and probabilistic ensemble forecasting. In practical terms, that means the model is designed not only to predict expected conditions, but also to show uncertainty across multiple forecast runs, a key capability for weather-sensitive decisions in energy, agriculture, transport, and climate-risk planning.

What changed

Aurora was first introduced as a research model for Earth-system prediction, and Microsoft says the new release extends that work with product engineering, cloud infrastructure, and managed access tied to Microsoft Weather services. The company is also releasing the model as open source on GitHub, with model checkpoints available through Hugging Face, so outside teams can evaluate and build on the system rather than treating it only as a closed commercial forecast service.

• New variables broaden the model beyond a narrower weather snapshot into more operationally relevant conditions.
• Hourly forecasts are intended to support decisions where daily or multi-hour averages are not detailed enough.
• Ensemble forecasting lets users compare likely outcomes and uncertainty instead of relying on a single deterministic run.

The release matters because weather foundation models are moving from research demonstrations toward practical infrastructure. Conventional numerical weather prediction remains central to forecasting, but AI models can be faster to run and easier to adapt when organizations need specialized views of atmospheric risk. Microsoft is positioning Aurora 1.5 as both an open research asset and a bridge into enterprise weather applications that need support, data pipelines, and operational assurance.

Microsoft's claims should still be tested against independent benchmarks and real-world deployments, especially for high-stakes uses such as disaster response or grid operations. But by publishing Aurora 1.5 openly and tying it to managed weather services, the company is giving the forecasting community a clearer path to compare AI weather models, adapt them to local needs, and assess where they can complement established meteorological systems.

Sources

• www.microsoft.com

Cover photo by Julien Goettelmann on Pexels, used under the Pexels License.