Nvidia is bringing the transformer architecture behind large language models (LLMs) to meteorology with two new open source models.

The New Stack is a publication covering trends and technologies in cloud-native development, DevOps, and software delivery. Developers can learn about containerization, Kubernetes, and cloud computing, as well as explore topics such as microservices architecture, serverless computing, and continuous integration/continuous delivery (CI/CD) pipelines.

The New Stack

Nvidia released two open source AI weather forecasting models, Earth-2 Medium Range and Earth-2 Nowcasting, using transformer architectures similar to large language models. The Medium Range model provides 15-day forecasts and reportedly outperforms DeepMind's GenCast across 70+ weather variables. The Nowcasting model generates country-scale forecasts at kilometer resolution. Nvidia also launched a Global Data Assimilation model that reduces the hours-long process of preparing initial weather data to seconds. These models aim to democratize weather forecasting beyond government supercomputers, making it accessible to enterprises for applications in agriculture, energy, aviation, and emergency response.

Nvidia makes AI weather forecasting more accessible, no supercomputer needed

Data assimilation: The other 50% of the problem