NVIDIA has unveiled its groundbreaking Earth-2 family of open models, marking a significant milestone in weather forecasting technology. The platform represents the world’s first fully open, accelerated set of models and tools specifically designed for AI-powered weather prediction, making professional-grade weather intelligence accessible to researchers, startups, and government agencies worldwide.
The Earth-2 platform includes several specialized models that address different forecasting needs. The Earth-2 Medium Range model, powered by the Atlas architecture, delivers highly accurate predictions for over 70 weather variables up to 15 days in advance, outperforming leading open models like GenCast on key industry benchmarks. Meanwhile, Earth-2 Nowcasting uses generative AI to predict satellite and radar imagery, providing high-fidelity, zero- to six-hour hazardous weather forecasts in minutes — becoming the first AI model to outperform traditional physics-based systems in simulating storm dynamics.
The real-world impact of Earth-2 is already being demonstrated by early adopters. The Israel Meteorological Service is using Earth-2 CorrDiff in operation and reports a remarkable 90% reduction in compute time at 2.5-kilometer resolution compared with running a classic numerical weather prediction model without AI. According to director Amir Givati, their AI model trained with CorrDiff was the best of all operational models for a six-hour verification of accumulated precipitation during a recent rainstorm.
Major organizations across multiple sectors are rapidly adopting Earth-2. In weather forecasting, AI tool provider Brightband is running Earth-2 Medium Range to issue real-world global forecasts daily, while The Weather Company is evaluating Earth-2 Nowcasting for localized severe-weather applications. The energy sector has shown particular interest, with TotalEnergies evaluating the platform to improve short-term risk awareness, and GCL, one of China’s largest solar material producers, deploying Earth-2 for its photovoltaic prediction system.
A key differentiator of Earth-2 is its fully open approach. Researchers and organizations can download Earth2Studio to start building, fine-tuning, and deploying Earth-2 open models on their own infrastructure. Julian Green, cofounder and CEO of Brightband, noted that the open-source nature of Earth-2 Medium Range speeds up innovation, allowing easier comparison and improvements by other members of the weather enterprise. With financial institutions like AXA and S&P Global also leveraging the platform for risk assessment, NVIDIA Earth-2 is positioning itself as the foundational technology for the next generation of weather and climate intelligence.
