Nvidia today released two open-source weather forecasting models, Earth-2 Medium Range and Earth-2 Nowcasting, plus a Global Data Assimilation tool to accelerate initial condition generation. The transformer-based Medium Range model predicts up to 15 days and reportedly outperforms DeepMind’s GenCast on over 70 variables, while Nowcasting provides kilometer-scale country forecasts adopted by services like Israel Meteorological Service. The assimilation model produces starting-state snapshots within seconds, easing compute bottlenecks.
Key Points
- 1Launches two transformer-based Earth-2 models and a global data-assimilation tool for forecasts
- 2Improves accuracy and resolution — Medium Range reportedly outperforms GenCast on 70+ variables
- 3Enables enterprise-level localized forecasts with faster initialization, reducing reliance on supercomputers
Scoring Rationale
High novelty and industry-wide applicability, but enterprise deployment requires significant data, compute resources and expertise.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
